[Senate Hearing 119-505]
[From the U.S. Government Publishing Office]
S. Hrg. 119-505
LESS HYPE, MORE HELP: AI THAT IMPROVES
SAFETY, PRODUCTIVITY, AND CARE
=======================================================================
HEARING
before the
SUBCOMMITTEE ON SCIENCE, MANUFACTURING,
AND COMPETITIVENESS
of the
COMMITTEE ON COMMERCE,
SCIENCE, AND TRANSPORTATION
UNITED STATES SENATE
ONE HUNDRED NINETEENTH CONGRESS
SECOND SESSION
__________
MARCH 3, 2026
__________
Printed for the use of the Committee on Commerce, Science, and Transportation
[GRAPHIC NOT AVAILABLE IN TIFF FORMAT]
Available online: http://www.govinfo.gov
______
U.S. GOVERNMENT PUBLISHING OFFICE
64-384 PDF WASHINGTON : 2026
SENATE COMMITTEE ON COMMERCE, SCIENCE, AND TRANSPORTATION
ONE HUNDRED NINETEENTH CONGRESS
SECOND SESSION
TED CRUZ, Texas, Chairman
JOHN THUNE, South Dakota MARIA CANTWELL, Washington,
ROGER WICKER, Mississippi Ranking
DEB FISCHER, Nebraska AMY KLOBUCHAR, Minnesota
JERRY MORAN, Kansas BRIAN SCHATZ, Hawaii
DAN SULLIVAN, Alaska EDWARD MARKEY, Massachusetts
MARSHA BLACKBURN, Tennessee GARY PETERS, Michigan
TODD YOUNG, Indiana TAMMY BALDWIN, Wisconsin
TED BUDD, North Carolina TAMMY DUCKWORTH, Illinois
ERIC SCHMITT, Missouri JACKY ROSEN, Nevada
JOHN CURTIS, Utah BEN RAY LUJAN, New Mexico
BERNIE MORENO, Ohio JOHN HICKENLOOPER, Colorado
TIM SHEEHY, Montana JOHN FETTERMAN, Pennsylvania
SHELLEY MOORE CAPITO, West Virginia ANDY KIM, New Jersey
CYNTHIA LUMMIS, Wyoming LISA BLUNT ROCHESTER, Delaware
Brad Grantz, Republican Staff Director
Nicole Christus, Republican Deputy Staff Director
Lila Harper Helms, Staff Director
Melissa Porter, Deputy Staff Director
------
SUBCOMMITTEE ON SCIENCE, MANUFACTURING,
AND COMPETITIVENESS
TED BUDD, North Carolina, Chairman TAMMY BALDWIN, Wisconsin, Ranking
MARSHA BLACKBURN, Tennessee GARY PETERS, Michigan
TODD YOUNG, Indiana JACKY ROSEN, Nevada
ERIC SCHMITT, Missouri JOHN HICKENLOOPER, Colorado
BERNIE MORENO, Ohio LISA BLUNT ROCHESTER, Delaware
CYNTHIA LUMMIS, Wyoming
C O N T E N T S
----------
Page
Hearing held on March 3, 2026.................................... 1
Statement of Senator Budd........................................ 1
Statement of Senator Baldwin..................................... 2
Statement of Senator Cruz........................................ 27
Statement of Senator Blunt Rochester............................. 30
Statement of Senator Blackburn................................... 32
Statement of Senator Moreno...................................... 34
Statement of Senator Hickenlooper................................ 36
Statement of Senator Young....................................... 38
Statement of Senator Cantwell.................................... 39
Statement of Senator Rosen....................................... 42
Witnesses
Demetri Giannikopoulos, Chief Innovation Officer, Rad AI......... 4
Prepared statement........................................... 5
Brittany Ng, Vice President, Siemens Digital Industries Software. 9
Prepared statement........................................... 10
Dr. Damion Shelton, Co-Founder and Chairman, Agility Robotics.... 14
Prepared statement........................................... 16
Mark Muro, Senior Fellow, Metropolitan Policy Program, Brookings
Institution.................................................... 17
Prepared statement........................................... 19
Appendix
Derek Monson, Executive Director, Sutherland Institute, prepared
statement...................................................... 47
Letter dated March 3, 2026 to Hon. Ted Budd and Hon. Tammy
Baldwin from Kristen Swearingen, Vice President, Government
Affairs, Associated Builders and Contractors................... 49
Letter dated March 3, 2026 to Hon. Ted Budd and Hon. Tammy
Baldwin from UVEye............................................. 50
Letter dated March 3, 2026 to Hon. Ted Budd and Hon. Tammy
Baldwin from Graham Dufault, General Counsel, ACT | The App
Association and Kedharnath Sankararaman, Policy Associate, ACT
| The App Association.......................................... 51
National Council on Disability, prepared statement............... 54
LESS HYPE, MORE HELP: AI THAT IMPROVES
SAFETY, PRODUCTIVITY, AND CARE
----------
TUESDAY, MARCH 3, 2026
U.S. Senate,
Subcommittee on Science, Manufacturing, and
Competitiveness,
Committee on Commerce, Science, and Transportation,
Washington, DC.
The Subcommittee met, pursuant to notice, at 10:15 a.m., in
room SR-253, Russell Senate Office Building, Hon. Ted Budd,
Chairman of the Subcommittee, presiding.
Present: Senators Budd [presiding], Cruz, Blackburn, Young,
Moreno, Baldwin, Cantwell, Hickenlooper, Blunt Rochester, and
Rosen.
OPENING STATEMENT OF HON. TED BUDD,
U.S. SENATOR FROM NORTH CAROLINA
Senator Budd. This hearing will come to order. Good
morning, everyone. Thank you all for being here.
I recognize myself for opening remarks, and thank you,
Ranking Member Baldwin and Chairman Cruz, Ranking Member
Cantwell, and our witnesses for working to put this important
hearing together.
Artificial intelligence will undoubtedly usher in
significant improvements in quality of life for the American
people. It will make many workplaces safer and more productive,
helping to increase output, raise wages, and grow the economy.
It will enhance manufacturing capabilities, especially
those of critical importance to our economic and national
security such as semiconductors and those in the defense
industrial base.
It will make it easier to reshore manufacturing through
human-enhancing automation and digital twinning simulation.
Smart systems and devices have the potential to
revolutionize health care, improving early detection of
diseases such as cancer, and helping people with disabilities
live better lives, not by replacing doctors but by augmenting
their diagnostic and treatment capabilities.
AI-enabled research, aided by self-driving cloud labs,
could massively reduce barriers to the discovery of new drugs.
AI's potential in the health care industry presents unique
opportunities to save and improve lives.
I have also noted that as I travel the state, some North
Carolinians share concerns about the growth of AI in automated
or autonomous technology.
There is a natural hesitancy toward technology that may
disrupt incumbent industries or systems. It is normal to worry
about the impacts of advancements on your job, your children,
and your community.
But if we have learned anything from our history, it is
that innovation is the lifeblood of the American economy. Our
nation's story has been shaped by technological advancements
that have time and time again expanded prosperity and improved
outcomes.
I believe that AI, deployed in numerous ingenious ways, can
help people be better versions of themselves in their daily
lives.
For those of us in this room, it is our job to listen to
those concerns and to work together in a bipartisan fashion to
address potential harms so that our Nation can reap the
tremendous benefits that AI has to offer.
As I have said before, winning the AI race against China is
paramount for our national and economic security. The
administration is leaning in and providing important
leadership, ongoing work to first identify and then to aid in
the export of the American AI stack.
From hardware to software, it will ensure that global AI
diffusion and standards are anchored in American values. The
Genesis Mission will build upon America's scientific dominance
by networking together world-class labs, computing power, and
datasets to turbo charge scientific discovery.
The diffusion of AI into the economy will be a critical
dimension in this race. According to one report, AI is the most
rapidly adopted general purpose technology in history with
three in five U.S. adults surveyed having used AI less than 3
years after its release.
However, other studies have found that U.S. businesses can
lag their Asian and even European peers in enterprise adoption
of AI tools.
I am concerned that China, given its top-down command-and-
control structure, deep and sophisticated manufacturing
industry, and open-source heavy AI ecosystem is in a prime
position to diffuse AI quickly and broadly.
Achieving a manufacturing renaissance in America is as
bipartisan and deeply held a goal as any in Congress. Given
demographic realities such as an aging, skilled workforce, and
stalled population growth, we will need to diffuse and scale
smart technologies and processes to make more critical goods
domestically.
I am excited to discuss many of these technologies and
systems today. Our witnesses are at the front lines and cutting
edge of making our economy and our daily lives smarter.
I look forward to hearing from them about what excites
them, what concerns them, and roadblocks we in Congress can
address.
Ranking Member, do you have any comments?
Senator Baldwin. Absolutely.
Senator Budd. All right.
STATEMENT OF HON. TAMMY BALDWIN,
U.S. SENATOR FROM WISCONSIN
Senator Baldwin. Thank you so much, Mr. Chairman, and thank
you to the witnesses who are appearing before our subcommittee
today.
Given the focus of today's hearing, I wanted to begin by
raising the troubling news from last week regarding the Trump
administration's actions to force artificial intelligence
companies to deploy systems without guardrails and safety
measures.
Anthropic refused to have their technology used for mass
surveillance of Americans or for the deployment of autonomous
weapons without human oversight. The Trump administration then
labeled Anthropic as a security risk, and Secretary Hegseth has
taken the unprecedented step to blacklist the company.
Our witnesses are here to testify about how artificial
intelligence can be used to improve worker safety and
efficiency, but we cannot ignore the stakes at play here.
Workers and consumers alike are rightfully worried about the
deployment of AI systems.
A study done by Stanford last year found that 45 percent of
workers expressed concerns about the accuracy and reliability
of AI systems. Twenty-three percent feared job displacement and
16 percent worried about a lack of human oversight built into
the systems.
And that does not even begin to touch on the concerns
regarding privacy, intellectual property, and energy use that
comes along with AI.
But if designed with workers at the table during all
stages, from research and development to deployment to
negotiating guardrails with employers, artificial intelligence
has the potential to improve productivity, reduce safety risks,
and ease burnout in the workplace.
The best innovations come from the factory floor. The
people actually doing the work know better than any senior
executive what the needs, inefficiency, and liabilities are day
to day.
So I look forward to hearing our witnesses today and,
again, thank you for being here.
Senator Budd. I thank the Ranking Member.
I would like to introduce our witnesses for the day. Our
first witness is Demetri--and help me on this one, OK?
Pronounce your last name for me.
Mr. Giannikopoulos. Giannikopoulos.
Senator Budd. I even have a phonetic spelling here, but
thank you for that.
Mr. Giannikopoulos is the Chief Innovation Officer at Rad
AI, a San Francisco-based company using AI to help radiologists
save time on reporting, allowing them to spend more time on
patient care. His experience applying new technologies to
health care spans over two decades.
Our second witness is Brittany Ng. Ms. Ng is Vice President
at Siemens Digital Innovative Industry Software. Ms. Ng is an
expert in business matters and she has extensive expertise in
the deployment of AI across manufacturing context.
Our third witness is Damion Shelton. Dr. Shelton is Co-
Founder and Chairman of the Board at Agility Robotics. He has
almost two decades of experience in academia, innovation, and
business, and Dr. Shelton founded two highly successful
robotics companies.
Our final witness is Mark Muro. Mr. Muro is a Senior Fellow
at the Metropolitan Policy Program at the Brookings
Institution. At Brookings, his research focuses on how
technology development plays out differently across regions.
All right. Mr. Giannikopoulos, you are recognized to
deliver your open statement if you are ready.
STATEMENT OF DEMETRI GIANNIKOPOULOS,
CHIEF INNOVATION OFFICER, RAD AI
Mr. Giannikopoulos. Chairman, Ranking Member, and members
of the Subcommittee, thank you for the opportunity to testify.
I come before you as a health care AI expert who has
deployed artificial intelligence nationwide, as a caregiver who
has supported my wife through cancer, as the spouse of a nurse
practitioner with more than a decade of frontline primary care
experience, and as a patient who has lived with multiple
sclerosis for more than two decades.
These experiences have shaped who I am and are why I am
here today. They have fueled my more than 20-year career in
health care. They have given me a firsthand view of how care is
delivered, how decisions are made, and what happens when the
system works and when it does not.
In each of these roles, I have seen the same reality.
Outcomes often hinge on whether a diagnosis is made quickly and
accurately, because in health care the most dangerous failure
is not a machine failure. It is a missed or delayed diagnosis.
Research from Johns Hopkins Medicine estimates that nearly
800,000 Americans each year die or are permanently disabled
because of diagnostic error. Extensive research has shown that
errors often occur because clinicians are operating within an
increasingly complex and strained health care system, managing
rising volumes, time pressures, and expanding information.
Imagine a patient arriving in a community emergency
department with sudden chest pain. A CT is performed. The cause
is something rare but deadly--an aortic dissection, a tear in
the body's main artery. Without rapid diagnosis and treatment,
a quarter of patients can die within 48 hours.
Today, AI is helping at multiple points in that patient's
care. Within minutes, AI can analyze the images and help flag
the findings so it is reviewed immediately.
When the radiologist opens the case, AI can help integrate
trusted clinical guidance directly into their workflow, drawing
on peer-reviewed evidence developed through organizations like
the Radiological Society of North America.
It does not replace the physician. It strengthens
competence and trust in the diagnosis. Once the diagnosis is
made, the care team must act quickly. The patient may need to
be transferred for life-saving surgery.
AI can help ensure the diagnosis is clearly communicated,
documented efficiently, and shared across care teams so
treatment can begin without delay.
That is productivity. That is coordinated care. Research
published by the American Heart Association shows that
coordinated care pathways for aortic dissection reduce delays
and improve survival.
We see the same reality in stroke care, where every minute
of delay risks permanent death of millions of brain cells,
controlling vital functions of the body. Faster diagnosis and
transfer can mean survival and less disability.
But the story does not end there. That same scan may reveal
a quiet lung mass--not an emergency today, but something that
must be followed carefully. AI can help ensure follow-up
imaging is scheduled, care teams are alerted, and patients do
not fall through the cracks months or years later.
Unfortunately, over half of patients never receive the
follow-up care their conditions require. For those Americans,
the risk is not theoretical. It becomes their reality.
That is, this is not just about efficiency. It is about
trust. We are asking clinicians to manage rising imaging
volumes, expanding documentation requirements, and increasingly
complex patient needs, while the health care workforce is
shrinking and burnout remains high.
AI, when implemented responsibly, is not replacing
clinicians. It is acting as a pressure release valve, helping
reduce cognitive burden and supporting clinicians in delivering
safe and timely care.
I have deployed both FDA-cleared AI tools and other
clinical software that does not require FDA clearance. In both
cases, these systems undergo extensive clinical, privacy, and
governance reviews before deployment.
Physicians remain responsible for patient care, and their
tools--these tools operate within existing health care laws and
professional accountability frameworks.
Based on my experience deploying these systems nationwide,
the most important determinant of safety is not only how they
are evaluated before deployment, but how they are governed,
monitored, and supported once they are in clinical use.
This is where thoughtful lifecycle governance and
consistent national standards are essential. This is especially
important for rural and underserved communities, where access
to subspecialty expertise may be limited.
AI can help extend the reach of clinical expertise and
support clinicians caring for patients, regardless of geography
and access to local resources. AI is most powerful not when it
replaces human judgment, but when it strengthens it.
AI will not replace physicians. It will help them do what
they train their entire lives to do, care for patients.
Less hype, more help--that is not a future promise. It is
happening today, because behind every scan is a person, a
family, and a moment where getting the diagnosis right can
change everything.
Thank you. I look forward to your questions.
[The prepared statement of Mr. Giannikopoulos follows:]
Prepared Statement of Demetri Giannikopoulos, Chief Innovation Officer,
Rad AI
Introduction
Chairman, Ranking Member, and Members of the Subcommittee:
Thank you for the opportunity to testify today on the role
artificial intelligence can play in strengthening healthcare for
patients and clinicians across the United States.
My name is Demetri Giannikopoulos, and I serve as Chief Innovation
Officer at Rad AI, where I oversee the development and responsible
clinical integration of artificial intelligence technologies used by
radiologists and health systems nationwide. My work focuses on ensuring
these systems are safely integrated into clinical workflows, support
physicians in delivering accurate and timely diagnoses, and help
strengthen care coordination and patient outcomes.
I previously served as Chief Transformation Officer at Aidoc, where
I led the clinical, operational, and governance protocol transformation
required to safely integrate multiple FDA-cleared artificial
intelligence medical devices into frontline physician workflows across
health systems nationwide. This work involved close collaboration with
clinicians, health system leadership, and regulatory stakeholders to
ensure these technologies functioned reliably in real-world clinical
environments and strengthened, rather than disrupted, the delivery of
patient care.
In parallel, I have been actively involved in national governance
efforts focused on establishing standards for responsible artificial
intelligence in healthcare. I participate in the Applied Model Card
Workgroup of the Coalition for Health AI, a clinician-led coalition
advancing transparency, accountability, and safety in healthcare
artificial intelligence. I also serve on the Artificial Intelligence
Accreditation Advisory Committee for URAC, where we are developing
accreditation frameworks to help ensure artificial intelligence is
implemented safely, consistently, and responsibly across diverse
healthcare settings.
I have also served on the Patient and Family Centered Care Clinical
Excellence Committee of the American College of Radiology, where I
contributed to the development of the Scanxiety Toolkit, which
addresses the anxiety patients experience while waiting for imaging
results. I bring this perspective not only through my professional
work, but also through personal experience. I supported my wife through
her cancer diagnosis and treatment, and I have also lived for more than
two decades as a patient with multiple sclerosis, undergoing regular
imaging and waiting for results that help determine the course of my
care. These experiences reinforced for me how deeply patients and
families depend on timely, accurate diagnosis, and clear communication
during some of the most vulnerable moments of their lives.
Together, these experiences have given me a comprehensive
perspective on both the extraordinary potential and the profound
responsibility associated with integrating artificial intelligence into
the practice of medicine. Artificial intelligence in healthcare has
reached an inflection point. The central question is no longer whether
these technologies are promising, but how they perform in real clinical
environments and whether they meaningfully improve safety,
productivity, and patient care.
Imaging as a Critical Component of Modern Diagnosis
Radiology plays a foundational role in modern medicine. Imaging is
often the step that confirms a diagnosis, rules out life-threatening
conditions, and determines the course of treatment. That matters
because diagnostic error remains a major source of preventable harm.
Research from Johns Hopkins Medicine, published in BMJ Quality &
Safety, estimates that each year nearly 800,000 Americans die or are
permanently disabled because of diagnostic error.\1\ Conditions such as
aortic dissection and stroke are among the exact types of time-
sensitive diagnoses that contribute to that harm.
---------------------------------------------------------------------------
\1\ David E. Newman-Toker et al., Burden of Serious Harms from
Diagnostic Error in the USA, BMJ Quality & Safety (2023), https://
pubmed.ncbi.nlm.nih.gov/37460118/.
---------------------------------------------------------------------------
Acute aortic dissection is a race against the clock. National
cardiology guidelines tell us that for dissections involving the
ascending aorta, the risk of death rises by about 1 to 2 percent every
hour without treatment.\2\ And we see the impact of that urgency very
clearly. Nearly one in four patients treated with medication alone died
within two days, compared to fewer than one in twenty patients who
received surgery.\3\
---------------------------------------------------------------------------
\2\ 2022 ACC/AHA Guideline for the Diagnosis and Management of
Aortic Disease (James de Lemos et al., eds., American College of
Cardiology/American Heart Association 2022) (noting that acute aortic
dissection of the ascending aorta is ``highly lethal in symptomatic
patients left untreated, with an early mortality of 1 percent to 2
percent per hour after symptom onset''), https://www.ahajournals.org/
doi/10.1161/CIR.0000000000001106.
\3\ Kevin M. Harris et al., Early Mortality in Type A Acute Aortic
Dissection: Insights From the International Registry of Acute Aortic
Dissection, JAMA Cardiology (2022), https://jamanetwork.com/journals/
jamacardiology/fullarticle/2795672.
---------------------------------------------------------------------------
Stroke care follows the same pattern. Nearly 1.9 million neurons
are lost each minute that treatment is delayed.\4\ In both cases, rapid
imaging is what makes timely treatment possible. It is what allows
physicians to see the problem, make the diagnosis, and act before
irreversible harm occurs. In these moments, imaging is not just
diagnostic. It is decisive.
---------------------------------------------------------------------------
\4\ Jeffrey L. Saver, Time Is Brain--Quantified, 37 Stroke 263
(2006), https://pubmed.ncbi.
nlm.nih.gov/16339467/.
---------------------------------------------------------------------------
Artificial intelligence can help strengthen this process. It can
flag life-threatening findings for rapid physician review, support
accurate interpretation, and help ensure that critical diagnoses are
communicated quickly so patients receive the care they need without
delay. This responsibility exists across all healthcare settings,
including rural and community hospitals where subspecialty expertise
may not always be immediately available. Artificial intelligence can
help reinforce this diagnostic infrastructure, supporting clinicians in
delivering timely and accurate care regardless of geography.
Artificial Intelligence Across the Diagnostic Continuum
Artificial intelligence supports care across the full diagnostic
continuum, from triage to interpretation to communication and follow-
up.
The first stage is triage and prioritization. Artificial
intelligence can analyze imaging studies shortly after acquisition and
help identify findings that require urgent attention. This helps ensure
that time-sensitive conditions are reviewed promptly, improving patient
safety.
The second stage is interpretation. Artificial intelligence can
help integrate trusted clinical knowledge and evidence directly into
physician workflows, supporting radiologists as they make diagnostic
decisions. These tools support clinicians, not replace them, and help
improve consistency and efficiency in care delivery. Artificial
intelligence can also improve clinical documentation and communication
by helping ensure diagnoses are clearly recorded, reports are completed
efficiently, and information is shared accurately across care teams.
This supports faster clinical action and reduces delays in treatment.
The third stage is follow-up and longitudinal care. Imaging
frequently identifies findings that require monitoring over time, yet
failure to complete recommended follow-up remains a significant patient
safety challenge. In one large academic health system study, only about
half of patients completed recommended follow-up imaging, with
completion rates of just 51.9 percent at one institution and 52.0
percent at another.\5\ Similarly, studies of pulmonary nodules have
shown that more than half of patients may not receive recommended
follow-up imaging.\6\
---------------------------------------------------------------------------
\5\ Thusitha Mabotuwana et al., Automated Tracking of Follow-Up
Imaging Recommendations, Am. J. Roentgenology (2019), https://
ajronline.org/doi/full/10.2214/AJR.18.20586.
\6\ Denitza P. Blagev et al., Follow-up of Incidental Pulmonary
Nodules and the Radiology Report, 11 J. Am. Coll. Radiology 378 (2014),
https://pubmed.ncbi.nlm.nih.gov/24316231/.
---------------------------------------------------------------------------
Artificial intelligence systems can help address this gap by
identifying patients who need follow-up, tracking whether that care
occurs, and supporting care teams in closing the loop. This helps
ensure that important findings do not fall through the cracks and that
patients receive the care they need. Together, these capabilities
improve safety, strengthen productivity, and support better care for
patients.
Workforce Challenges and Cognitive Burden
Healthcare systems today face growing workforce challenges. Imaging
volume has increased substantially over time, while workforce growth
has not kept pace with demand.\7\ This imbalance reflects broader
trends in healthcare, driven by population aging and expanded reliance
on diagnostic imaging. At the same time, clinician exhaustion,
cognitive overload, and burnout remains widespread, affecting a
substantial portion of the clinical workforce.\8\
---------------------------------------------------------------------------
\7\ Eric C. Christensen et al., Projecting the Future Radiologist
Workforce Supply in the United States Through 2055, 21 J. Am. Coll.
Radiology (2024); Eric C. Christensen et al., Projecting Imaging
Utilization in the United States Through 2055, 21 J. Am. Coll.
Radiology (2024)., https://www.jacr.org/article/S1546-1440(24)00898-6/
fulltext, https://www.jacr.org/article/S15
46-1440(24)00909-8/fulltext.
\8\ Tait D. Shanafelt et al., Changes in Burnout and Satisfaction
With Work-Life Integration in Physicians and the General U.S. Working
Population Between 2011 and 2023, 100 Mayo Clinic Proc. 1142 (2025),
https://www.mayoclinicproceedings.org/article/S0025-6196(24)00668-2/
fulltext.
---------------------------------------------------------------------------
Artificial intelligence can help address this challenge by
improving productivity, reducing administrative burden, and supporting
clinicians as they manage increasing workload demands. A recent task-
based analysis by Dr. Curtis Langlotz of Stanford University, the
immediate past President of the Radiological Society of North America,
projects that artificial intelligence could significantly reduce the
time radiologists spend on certain tasks, particularly report drafting
and workflow coordination. However, the study also concludes that
because imaging demand continues to grow and the radiology workforce
has remained relatively stable, artificial intelligence is unlikely to
eliminate the need for radiologists.\9\ Instead, these technologies are
expected to help clinicians work more efficiently and focus more of
their time on direct patient care.
---------------------------------------------------------------------------
\9\ Langlotz CP. The Effect of AI on the Radiologist Workforce: A
Task-Based Analysis. medRxiv [preprint]. Posted December 22, 2025.
doi:10.64898/2025.12.20.25342714, https://www.medr
xiv.org/content/10.64898/2025.12.20.25342714v1.
---------------------------------------------------------------------------
In addition to improving efficiency, artificial intelligence can
help reduce cognitive burden by assisting with information management,
documentation, and follow-up tracking. This support helps clinicians
focus their attention on patient care and clinical decision making,
which is especially important as workforce shortages and burnout
continue to affect healthcare nationwide. Artificial intelligence
supports clinicians; it does not replace them.
Improving Access to Care
Healthcare access challenges are particularly acute in rural and
underserved communities. More than 46 million Americans live in rural
areas, where access to specialists may be limited.\10\ Many rural
hospitals operate with constrained clinical staffing and may not have
immediate access to subspecialty expertise, including radiologists with
advanced training in specific conditions.
---------------------------------------------------------------------------
\10\ U.S. Census Bureau, New Census Data Show Differences Between
Urban and Rural Populations (2016), https://www.census.gov/newsroom/
archives/2016-pr/cb16-210.html.
---------------------------------------------------------------------------
Artificial intelligence can help address these gaps by supporting
clinicians in delivering timely and effective care. For example,
artificial intelligence can help identify urgent findings, assist with
interpretation, and ensure that critical results are communicated
promptly. These capabilities help clinicians work more efficiently and
reduce the risk that important findings are delayed or missed. Ensuring
that patients receive consistent, high-quality diagnostic care
regardless of geography is one of the most important opportunities for
artificial intelligence to improve healthcare nationally.
Governance, Oversight, and Accountability
The safe and effective use of artificial intelligence requires
strong governance, clear accountability, and ongoing oversight.
Based on my experience deploying these systems nationwide, the most
important determinant of safety is not only how artificial intelligence
is evaluated before deployment, but how it is governed, monitored, and
supported once it is in clinical use. Importantly, artificial
intelligence systems operate within existing legal and professional
accountability frameworks. Physicians remain responsible for clinical
decision making, and health systems are responsible for safe
implementation. Existing laws, including patient privacy protections,
medical malpractice standards, and civil rights protections, continue
to apply. Artificial intelligence does not replace these safeguards; it
operates within them.
Governance initiatives such as those led by the Coalition for
Health AI and accreditation programs such as those being released by
URAC play an important role in supporting safe and responsible
implementation. Maintaining clinician leadership in patient care
decisions remains essential to ensuring patient safety and preserving
trust.
Regulatory Considerations and Responsible Innovation
The United States has established strong regulatory frameworks to
evaluate medical devices and protect patient safety. Through my
experience implementing FDA-cleared artificial intelligence systems, I
have seen how regulatory clarity and predictability support responsible
innovation and safe deployment. Peer-reviewed research has consistently
shown that artificial intelligence performs best when used to support
clinicians, not replace them.\11\
---------------------------------------------------------------------------
\11\ Eric J. Topol, High-Performance Medicine: The Convergence of
Human and Artificial Intelligence, Nature Medicine (2019), https://
www.nature.com/articles/s41591-018-0300-7.
---------------------------------------------------------------------------
Artificial intelligence introduces new considerations because these
technologies operate within complex clinical environments and continue
to evolve over time. Ongoing collaboration among regulators, healthcare
providers, and developers can help ensure regulatory frameworks
continue to support patient safety while enabling beneficial
innovation. Thoughtful evolution of regulatory approaches can
strengthen both safety and public trust.
Conclusion
Artificial intelligence is already helping strengthen healthcare
delivery when it is implemented responsibly, governed carefully, and
designed to support clinicians and patients. These technologies can
help clinicians diagnose disease earlier, improve productivity,
strengthen care coordination, and improve patient outcomes.
Artificial intelligence supports clinicians; it does not replace
them. It helps clinicians do what they trained their entire lives to
do: care for patients. Behind every scan is a patient, a family, and a
moment where getting the diagnosis right can change the course of a
life. Thank you for the opportunity to testify. I look forward to your
questions.
Senator Budd. Thank you for your remarks.
Ms. Ng, you are recognized for 5 minutes.
STATEMENT OF BRITTANY NG, VICE PRESIDENT,
SIEMENS DIGITAL INDUSTRIES SOFTWARE
Ms. Ng. Good morning, Chairman, Ranking Member, and members
of the Subcommittee. Thank you for the opportunity to testify
today.
The United States does not need AI as an abstract
capability. Our global competitiveness requires AI deployed on
factory floors, in shipyards, and across production systems
where it generates measurable productivity gains.
I am proud to take on this challenge. I lead Siemens'
maritime business in the United States where we work hand in
hand with shipyards and manufacturers.
At Siemens, we are a global leader in industrial AI. Last
year alone we invested $15 billion in the United States to
further our leadership in this transformational technology.
We are continuing to bring industrial AI to customers and
seeing firsthand how AI is most powerful when it connects data
to operational decisions in the real world.
Through our maritime work, shipbuilders are using our
physics-based, AI-enabled digital shipbuilding platform to
connect, design, simulation, and production planning so that
teams can identify bottlenecks before they occur.
They can reduce rework, improve first-time quality, and
compress production timelines, and by training AI in virtual
environments, shipyards can improve planning and performance
even in complex production conditions.
Industrial AI, alongside digital twins and software-defined
automation, is transforming manufacturing across industries. We
are seeing machine downtime reduced by 50 percent, energy
consumption cut by 20 percent, and quality control with 99.99
percent accuracy.
Industrial AI is helping manufacturers to achieve up to 40
percent in productivity improvements and it is transforming
traditional factories into flexible digital enterprises.
In modern shipbuilding, innovation looks like being able to
create a full digital twin of a vessel before steel is even
cut. This means that our customers can simulate how the ship
will be built, how systems will integrate, and how production
will flow through the yard, all in a virtual environment.
On the deck plate, industrial AI can help sequence work
packages, flag quality issues earlier, predict equipment
downtime, and optimize the flow of material. The crews
deploying these technologies are spending less time waiting and
doing rework and more time building.
When these tools are broadly adopted, they become economic
multipliers. However, technology does not just transform
industry unless the people understand it, trust it, and see how
it improves their lives.
At Siemens, we have learned that digital transformation
works best when the worker is at the center. Manufacturing
today faces a shortage of 400,000 open jobs nationwide.
The issue is not AI coming for industrial jobs. It is a
shortage of skilled workers amid increasing production
complexity. Industrial AI does not just solve the workforce
shortage; it changes the entire equation.
These technologies reduce repetitive and hazardous tasks
while elevating the skill sets of American workers. As Congress
considers AI and manufacturing policy, I respectfully offer
three considerations.
First, distinguish industrial AI from consumer AI.
Industrial AI operates in structured, safety-critical,
business-to-business environments with validated data and
rigorous testing.
Second, focus on adoption. Federal policies and investments
should integrate digital modernization from the outset and
accelerate deployment through public-private collaboration.
Programs like Manufacturing USA and the DOE's Genesis
Mission provide that bridge, moving from innovation into real-
world production.
Last, the government should send a clear and consistent
signal that digital capability is core industrial
infrastructure. America's competitive edge will be determined
by who most effectively deploys AI in the physical world.
Our nation is indisputably the world's greatest innovation
leader. Our challenge is scaling advanced technologies through
the industrial base.
Thank you, and I look forward to your questions.
[The prepared statement of Ms. Ng follows:]
Prepared Statement of Brittany Ng, Vice President,
Siemens Digital Industries Software
Introduction
Chairman Cruz, Ranking Member Cantwell, Chairman Budd, Ranking
Member Baldwin, and distinguished Members of the Subcommittee, thank
you for the opportunity to testify today.
I lead Siemens' maritime business in the United States, where we
work directly with shipyards, defense contractors, and manufacturers
modernizing some of the most complex production systems in the world.
Siemens is a global technology company with deep roots in the
United States, including the global headquarters of our software
business in Plano, TX. Across our businesses, we help design, simulate,
build, and operate the infrastructure and industrial systems that
underpin the American economy. Last year alone, Siemens invested
approximately $15 billion in the U.S. to deepen our leadership in
Industrial AI, simulation, and digital engineering.
Through that work, we have learned a simple lesson: AI is most
powerful not as a standalone capability, but when it connects data to
operational decisions in the physical world.
Siemens in the USA
Siemens has been part of the American industrial landscape for over
160 years. Our presence in the United States spans manufacturing
floors, shipyards, software labs, and engineering classrooms. In many
ways, our technologies are embedded in nearly every chapter of the
American industrial story--from designing semiconductors and building
aircraft, to modernizing shipyards, powering data centers,
strengthening the electric grid, and advancing next-generation
manufacturing. Approximately 50,000 Siemens employees across the
country are designing, building, and deploying the technologies that
power American production.
In Fort Worth, Texas, one of our newest advanced manufacturing
facilities illustrates what that commitment looks like in practice.
Before the first piece of equipment was installed, the factory existed
as a full digital twin--a physics-based virtual replica used to
simulate production flows, validate line configurations, and stress-
test material movement. By resolving bottlenecks in the virtual
environment first, the facility ramped up faster with greater precision
and flexibility. It is a clear example of how digital-first planning
accelerates real-world industrial capacity.
That same approach extends beyond a single facility. Across the
Midwest--from Missouri to Ohio--thousands of Siemens engineers and
software developers build the digital engineering platforms that make
these deployments possible. The simulation and industrial software
developed by these teams form the backbone of AI-enabled production
systems used across the United States and globally. Industrial AI
adoption is not just happening on factory floors--it is being coded,
tested, and refined in American software hubs.
Building on this momentum, Siemens is also advancing an AI build-
out in Washington state to scale AI and digital infrastructure where
industrial transformation is accelerating. The goal of our Data and AI
team, which will be globally headquartered in the United States, is to
unify data, software, and engineering platforms to deliver AI-native
solutions across the industrial lifecycle, from design through
operations. Coupled with Siemens' broader U.S. investment strategy--
totaling investments of over $100 billion in 20 years--this build-out
reinforces our belief that scaling Industrial AI is essential to
competitiveness.
Our commitment to the United States is not limited to facilities
and software--it is equally a commitment to people. Siemens has pledged
to help train 200,000 electricians and manufacturing experts in the
United States by 2030 to close the skills gap that increasingly defines
industrial competitiveness. This initiative reflects a simple reality:
as production systems become more software-defined and AI-enabled, the
competitive advantage of the United States will depend not just on
innovation, but on whether America's workforce is prepared to lead that
innovation on the factory floor, in the software suite, and at the
intersection of design and execution.
That commitment comes to life through deep partnerships across the
country with universities, apprenticeships, vocational schools, and
other workforce institutions preparing the next generation of
industrial leaders. At Purdue University, Siemens collaborates to
advance digital twin research and embed simulation tools directly into
engineering education--ensuring graduates are fluent in the digital
environments they will encounter in modern manufacturing. In Colorado,
Siemens drives microcredentialing initiatives that provide modular,
industry-recognized digital skills training. And in Kansas, Siemens
technologies power smart factory initiatives that demonstrate how AI-
enabled automation can scale across manufacturers of all sizes.
The list can go on. But, all of our efforts reflect a consistent
theme: Siemens' role in the United States is operational, not abstract.
We design here. We manufacture here. We train the workforce here. And
we deploy Industrial AI here--in ways that strengthen American
productivity and industrial readiness.
Industrial AI in the Physical World
In modern shipbuilding, we can now create a full digital twin of a
vessel before steel is ever cut. A digital twin is a high-fidelity,
physics-based virtual demonstration of a physical asset or production
process. In shipbuilding, that means we can simulate how the ship will
be built, how systems will integrate, how materials will flow through
the yard, how production constraints may emerge--all before physical
construction begins.
For example, through our work with HD Hyundai, shipbuilders are
deploying an AI-enabled digital shipbuilding platform that connects
design, simulation, and production planning. Teams can test decisions
virtually, identify bottlenecks before they occur, reduce costly
rework, and compress production timelines.
By training AI in virtual environments using digital twins and
synthetic data, shipyards can improve real-world planning and
performance in highly complex production environments. This is what AI
looks like in the physical world.
On the deck plate, it means sequencing work packages more
efficiently, flagging quality issues earlier, predicting equipment
downtime, and optimizing material movement across congested yards.
Workers also get to spend less time waiting and correcting mistakes,
and more time building.
AI also offers tangible improvements for worker safety in
manufacturing, as systems can identify hazardous conditions before they
escalate--whether that is equipment operating outside safe thresholds,
congestion in high-traffic work zones, or potential conflicts between
tasks. Digital twins allow crews to simulate complex lifts, confined-
space operations, and workflow changes virtually before executing them
in the physical environment. Workers spend less time navigating
uncertainty and fewer hours exposed to avoidable risks. Instead of
reacting to problems after they occur, teams can anticipate them--
improving both safety outcomes and operational confidence.
What is true in maritime manufacturing is true across sectors.
Industrial AI and software-defined automation are delivering measurable
gains across aerospace, automotive, semiconductors, energy systems,
food & beverage, advanced manufacturing, and more.
We're seeing machine downtime reduced by 50 percent, energy
consumption cut by 20 percent, and quality control with 99.99 percent
accuracy across industrial operations. Furthermore, Industrial AI
capabilities are helping manufacturers achieve up to 40 percent
productivity improvements through predictive maintenance, quality
control optimization, and automated decision-making--transforming
traditional factories into flexible, efficient digital enterprises.
For example, through our partnership with JetZero--an American
aerospace innovator--we are helping bring to market a next-generation
blended-wing aircraft and a digitally native ``Factory of the Future''
in North Carolina. Using Siemens' AI-enabled digital twin and
engineering platforms, JetZero can explore thousands of design
configurations, model aerodynamic performance, and simulate its
production system before physical assembly begins. AI tools support
both engineering and shopfloor operations--from optimizing aircraft
performance to assisting technicians in diagnosing equipment issues.
The result is a more efficient aircraft and a smarter manufacturing
system built here in the United States.
Across all these sectors--the result of this adoption is faster
time to market, reduced waste, lower cost structures, and more
resilient supply chains.
AI is not an abstract capability. AI is deployed on factory floors,
in shipyards, and across production systems where it generates
measurable productivity gains.
Accelerating Adoption
America's competitive edge will not be determined by who builds the
most AI models. It will be determined by who most effectively deploys
AI in the physical world.
That requires adoption.
The United States does not have an innovation problem--we have an
adoption challenge. We lead in developing breakthrough technologies.
The strategic challenge is scaling them across manufacturers of all
sizes and across every sector of our industrial base.
This adoption challenge is particularly acute for small and medium-
sized manufacturers, which often lack the ability to experiment with
emerging technologies. Through our work with Haddy--an American
advanced additive manufacturing company--we are demonstrating how
automation-first production can lower those barriers. Haddy operates
localized micro-factories that combine large-scale robotic 3D printing
with digitally coordinated workflows. AI is embedded directly into
factory control and execution systems, enabling faster iteration and
scalable distributed production. This model shows how smaller
manufacturers can compete in advanced industrial supply chains without
the footprint of traditional larger-scale facilities.
This is also true in national security contexts. In shipbuilding,
small improvements in production or quality can compound across
thousands of tasks and millions of labor hours. Adoption translates
directly into more predictable schedules, lower costs, and stronger
industrial capacity.
At Siemens, we create technology to transform the everyday, for
everyone. But technology alone does not transform an industry.
Deployment at scale does. The companies and nations that integrate AI
into core production systems--not as isolated pilots, but as operating
infrastructure--will lead in productivity and competitiveness.
AI & The Manufacturing Workforce
Technology cannot transform industry unless people trust it,
understand it, and see how it improves their work. At Siemens, we view
AI as a people-centric transformation designed to help the workforce
make better decisions at a faster pace, rather than replace human
expertise.
Manufacturing today faces more than 400,000 open jobs nationwide.
The issue is a shortage of workers combined with increasing production
complexity. Across the country, an aging workforce is retiring faster
than it can be replaced, and many employers struggle to attract and
retain the next generation of digitally native talent. Young workers
expect modern tools, intuitive systems, and career mobility--not paper-
driven processes and repetitive manual tasks.
Industrial AI doesn't just solve the workforce shortage. It changes
the equation.
Digital twins are a force-multiplier and allow workers to train in
simulated environments before stepping onto the production floor. AI-
powered copilots help technicians diagnose issues faster. Software-
defined systems make production more intuitive and adaptable. Across
environments, from the deck plate to the factory floor, this means
fewer repetitive and hazardous tasks, faster troubleshooting, and more
predictable execution.
These technologies elevate the skill sets of American workers for
the higher paying jobs of tomorrow, while also increasing productivity
per worker. At Siemens, we've learned digital transformation works best
when the worker is at the center--when welders, machinists, planners,
and engineers are part of the process from day one.
Industrial AI empowers workers to operate more advanced and
competitive ecosystems--thereby strengthening both productivity and
opportunity.
Policy Considerations
As Congress considers AI and manufacturing policy, we offer three
considerations.
1. Distinguish Industrial AI from Consumer AI
Industrial AI operates in structured, safety-critical,
business-to-business environments using validated operational
data generated by machines, sensors, and engineering systems.
These applications are deployed in regulated industries such as
shipbuilding, aerospace, energy, and semiconductor
manufacturing, where safety and quality are non-negotiable.
Industrial AI applications often occur upstream in design and
simulation environments before physical products are built.
Digital twins allow manufacturers to test and optimize
processes virtually, reducing risk before deployment in the
real world. In these settings, AI systems operate within
defined parameters and controlled data environments--very
different from open-ended consumer applications.
To foster innovation, policy frameworks should reflect these
distinctions. Approaches designed for consumer-facing AI should
not unintentionally slow adoption in manufacturing environments
that already operate under rigorous technical and safety
standards.
2. Focus on Adoption
The United States faces adoption and scaling challenges,
especially as emerging technologies are being deployed
worldwide at a rapid pace. To compete in the global economy,
breakthrough technologies developed in labs and leading firms
must be adopted broadly across the industrial base--including
small and medium-sized manufacturers--to drive economy-wide
productivity gains.
In tandem, Federal policies and investments--from shipbuilding
to semiconductors--must integrate digital tools from the outset
to ensure advanced manufacturing initiatives include the
technology necessary for AI-enabled production.
Public-private collaboration plays a critical role in
accelerating this deployment. Programs that connect advanced
computing, simulation, and industrial software with real-world
production environments can shorten commercialization timelines
and reduce barriers to adoption. Through the Department of
Energy's Genesis Mission, Siemens is excited to help translate
Industrial AI, digital twins, and advanced computing into
scalable, operational industrial applications.
Innovation creates meaningful potential. Adoption converts that
potential into measurable productivity, resilience, and
competitiveness.
3. Send a Clear Signal for Digitally Native Production
Both industry and government should send clear and consistent
signals that digital tools are core infrastructure for
competitiveness.
For example, Manufacturing USA Institutes provide platforms
where manufacturers, suppliers, technology providers, and
workforce institutions collaborate on applied innovation.
Managing these partnerships as a true network to focus on
specific manufacturing challenges can help pursue the goal of
bridging the gap between basic research and commercialization.
Tailored regulatory sandboxes complement this effort by
proposing structured environments where AI-enabled production
systems can be deployed and tested within defined guardrails.
This approach allows responsible experimentation to move
forward with clarity and confidence, putting technology and
innovation before regulation.
Additionally, challenge-based partnerships such as the
Department of Energy's Genesis Mission help translate
breakthrough capability development into operational systems
that can create real-world impact. By connecting advanced
research infrastructure to real use cases, these programs
bridge the gap between innovation and deployment.
Modernization scales fastest when expectations are clearly set
in Federal programs, such as procurement standards, financing
mechanisms, workforce programs, and public-private
partnerships. A coordinated signal from policymakers that
digitally native industrial strategy is a national priority
will accelerate technology adoption, strengthen domestic supply
chains, and reinforce long-term industrial readiness.
Closing
In closing, consider the story about Albert Einstein riding on a
train. When the conductor came by to collect tickets, Einstein began
searching his pockets but couldn't find his ticket. The conductor
recognized him and said, ``Professor Einstein, it's okay--I know who
you are.''
Einstein kept searching. He looked in his briefcase, under his
seat, even on the floor. The conductor said again, ``Dr. Einstein, I
trust you.''
Einstein replied, ``Young man, I know who I am. What I don't know
is where I'm going.''
America knows who it is: the world's great innovation leader. But
if we don't focus on where we're going--on how we deploy these
groundbreaking innovations across our industrial base--we risk losing
direction.
If we lead in adoption, we lead in productivity.
If we lead in productivity, we lead in competitiveness.
Thank you, and I look forward to your questions.
Senator Budd. Dr. Shelton, you are recognized for 5
minutes.
STATEMENT OF DR. DAMION SHELTON, CO-FOUNDER
AND CHAIRMAN, AGILITY ROBOTICS
Mr. Shelton. Thank you, Chairman Budd and Ranking Member
Baldwin, for inviting me to speak today. Also, thanks to the
rest of the Committee.
I would like to open with a quote from ``Peter Pan''--all
of this has happened before and it will all happen again.
The United States first started collecting census data
around agricultural employment in 1820, almost exactly 200
years before Agility Robotics launched its humanoid robot.
In 1820, the vast majority of Americans worked on farms
using simple tools and assisted by animals. Over the next 200
years, agricultural employment shrank to about 2 percent of the
workforce.
This is not, however, a story of decline. Absolute
employment, the total number of Americans working in
agriculture, has actually increased by about 30 percent since
the 1820s.
Why? In a single word, technology. In the 50 years after
the patenting of the McCormick reaper in 1834, nearly 12,000
additional farm implement patents were filed in the U.S. Far
from destroying jobs, automation in agriculture expanded both
direct agricultural employment and unlocked truly staggering
growth in the United States, a 30X increase in population and a
40X increase in per capita GDP.
The modern AI and robotics boom promises to bring these
sorts of transformative changes to the rest of our economy.
Agility Robotics started in 2015 as a spinoff from Oregon State
University.
From our DARPA-funded academic roots we have grown to more
than 300 employees, primarily in Oregon, Pennsylvania, and
California. Our Oregon facility co-locates both our R&D teams
and our factory, RoboFab, where we assemble 100 percent of our
robots in the U.S.
Our humanoid robot, Digit, was launched in 2020 as the
first full-size humanoid robot available for purchase. Our
customers include warehouse and logistics work, Amazon, GXO,
and Mercado Libre, and manufacturing and automotive suppliers
Schaeffler and Toyota with the common theme of providing a
solution for repetitive material handling alongside human co-
workers.
I would like to address the impact of automation on jobs
and human workers. Many manual labor jobs in the U.S. are
facing pressure from an aging workforce, high turnover, and a
high injury rate, which is where Agility has focused its
deployments.
The Bureau of Labor Statistics projects average annual
vacancies of more than one million positions in warehouse and
logistics work. Each robot deployed here not only does not take
a job from a human, it enables the business to grow and expand
overall hiring elsewhere in the organization.
To cite one example from an Agility partner, in 2012,
Amazon employed about 88,000 people and acquired the robotics
startup Kiva. Over the next 13 years, they deployed more than a
million robots while also hiring an additional 1.4 million
people.
Prior to the modern AI boom, it took a skilled engineer
many months to develop a new application for a robot. Modern AI
can dramatically shorten the time to deployment for small and
medium-sized businesses that desperately need to solve labor
challenges but lack the technical and financial resources to
run their own IT department.
However, as we have seen from self-driving vehicles, safety
often lags behind technical ability. It is imperative that
robots not endanger their human colleagues, members of the
public, or our homes and workplaces.
Focusing first on controlled environments has allowed
Agility Robotics and our partners like Boston Dynamics to focus
on developing responsible industry-led safety standards for
robots in the workplace.
Gaining experience in these environments first with a focus
on continuous safety improvement and industrial best practices
is the fastest and most responsible path toward a successful
long-term buildout of general purpose automation.
As an industry, we owe the general public a solid safety
argument backed by data. I will conclude by briefly addressing
the challenges posed by China's extremely rapid progress in
humanoid robots.
To be as blunt as possible, they are doing a good job.
Their technology is well-designed, highly capable, and backed
by a formidable supply chain. Two recent papers by security
researchers have identified a critical security vulnerability
in a Chinese humanoid robot allowing for remote takeover as
well as a phone home data logging mechanism that sends data to
a remote server.
As of this week--and I did check this before the hearing--
that robot is available for sale in the U.S. for less than
$20,000. Combined with open source and open weight AI models,
China is creating a compelling value proposition for early
adopters of general purpose humanoid robots.
This is a siren's call that we would be well-served to take
seriously or risk ceding the future of automation and, by
extension, economic growth to others.
Thank you again for the opportunity to speak today and I
thank for this committee's focus and leadership on these
important issues, including the regulatory sandbox approach.
Agility Robotics is ready to help the U.S. solve its labor
challenges, expand the economy, and ensure American
competitiveness and economic security as we move into the next
industrial revolution.
[The prepared statement of Mr. Shelton follows:]
Prepared Statement of Dr. Damion Shelton, Co-Founder and Chairman of
the Board, Agility Robotics
Thank you Chairman Budd and Ranking Member Baldwin for inviting me
to speak to you today. And thank you as well to Full Committee Chairman
Cruz and Ranking Member Cantwell, and all the other Members of this
subcommittee.
I'd like to open with a quote from Peter Pan: ``All this happened
before, and it will happen again.''
The United States first started collecting census data on
agricultural employment in 1820, almost exactly 200 years before
Agility Robotics launched its humanoid robot. In 1820, the vast
majority of Americans worked on farms, using simple tools and assisted
by animals. Over the next 200 years, agriculture employment shrank to
about 2 percent of the overall workforce. This is not, however, a story
of decline. Absolute employment--the number of total Americans working
in agriculture--has actually increased by almost 30 percent since the
1820s. Why? In a single word: technology. In the 50 years after the
patenting of the McCormick Reaper in 1834, nearly 12,000 additional
farm implement patents were filed. Far from destroying jobs, automation
in agriculture expanded both direct agricultural employment, and
unlocked truly staggering growth in the United States: a 30x increase
in population and a 40x increase in per-capita GDP. The modern AI and
robotics boom promises to bring these sorts of transformative changes
to the rest of our economy.
Agility Robotics was started in 2015, as a spin-off from Oregon
State University. From our DARPA-funded academic roots, we've grown to
more than 300 employees primarily in Oregon, Pennsylvania, and
California. Our Oregon facility co-locates both our R&D teams and our
factory--RoboFab, and 100 percent of our robots are assembled in the
USA. Our humanoid robot, Digit, was launched in 2020 as the first full-
size humanoid robot available for purchase. Our customers include
warehouse and logistics work (Amazon, GXO, and Mercado Libre) and
manufacturing and automotive (Schaeffler and Toyota), with the common
theme of providing a solution for repetitive material handling tasks
alongside human coworkers.
I'd like to address the impact of automation on jobs and human
workers. Many manual labor jobs in the U.S. are facing pressure from an
aging workforce, high turnover, and a high injury rate, which is where
Agility has focused its deployments. The Bureau of Labor Statistics
projects average annual vacancies of more than one million positions in
warehouse manual labor roles.
Each robot deployed here not only doesn't take a job from a human,
it enables the business to grow and expand overall hiring elsewhere in
the organization. To cite one example from an Agility partner: in 2012,
Amazon employed about 88,000 people and acquired the robotics startup
Kiva. Over the next 13 years, they deployed more than a million robots,
while hiring an additional 1.4 million people.
Prior to the modern AI boom, it took a skilled engineer many months
to develop a new application for the robot. Modern AI can dramatically
shorten the time to deployment for the small and medium-sized
businesses that desperately need to solve labor challenges but lack the
technical and financial resources to run their own IT department.
However, as we've seen with self-driving vehicles, safety often lags
behind technical ability. It's imperative that robots not endanger
their human colleagues, members of the public, or our homes and
workplaces.
Focusing first on controlled environments has allowed Agility
Robotics and our partners like Boston Dynamics to focus on developing
responsible, industry-led safety standards for robots in the workplace.
Gaining experience in these environments first, with a focus on
continuous safety improvement and industrial best practices, is the
fastest and most responsible path towards a successful long-term
buildout of general-purpose automation. As an industry, we owe the
general public a solid safety argument backed by data.
I'll conclude by briefly addressing the challenges posed by China's
extremely rapid progress in humanoid robotics. To be as blunt as
possible: they're doing a good job. Their technology is well designed,
highly capable, and backed by a formidable supply chain. Two recent
papers by security researchers have identified a critical security
vulnerability in a Chinese humanoid robot allowing for remote takeover,
as well as a ``phone home'' data logging mechanism that sends
continuous operational data to a remote server. As of this week, that
robot is available for sale in the U.S. for less than $20,000. Combined
with open source and open weight AI models, China is creating a
compelling value proposition for early adopters of general-purpose
humanoid robots. This is a siren's call that we would be well served to
take seriously, or risk ceding the future of automation, and by
extension continued economic growth, to others.
Thank you again for the opportunity to speak today, and I thank you
for this committee's focus and leadership on these important issues,
including the regulatory sandbox approach. Agility Robotics is ready to
help the U.S. solve its labor challenges, expand the economy, and
ensure American competitiveness and economic security as we move into
the next industrial revolution.
Senator Budd. Thank you, Dr. Shelton.
Mr. Muro, you are recognized for 5 minutes.
STATEMENT OF MARK MURO, SENIOR FELLOW,
METROPOLITAN POLICY PROGRAM,
BROOKINGS INSTITUTION
Mr. Muro. Mr. Budd, Ms. Baldwin, thanks so much. And to the
distinguished members of the Committee, I really want to thank
you for the opportunity to comment here.
I will just say my remarks here are my personal views, you
know, separate from Brookings Institution, and I think I want
to start by reaffirming the basic premise here that I think is
very compelling, that America's innovators are day by day
introducing incredible new tools and solutions that are
allowing more and more people, firms, and entrepreneurs to
expand and reach--reach and achievement of human expertise.
It is a compelling moment, and that is creating in the
country significant excitement, and yet a degree of pessimism
also complicates this moment. We should be frank about that and
the Committee has been frank about that.
Some worry that the technology will not be smoothly
adopted, that adoption will be challenging. Others fear what
has been deemed the greatest automation technology in history,
and still others worry about safety questions and which skills
will serve them in the future to enable this work.
So I want to say just a few words about areas where Federal
support can help maintain the sector's momentum, reinforce its
value, and promote optimism about its possibilities.
In this direction, my fellow panelists have done an
excellent job of inspiring confidence by detailing some of the
possibilities. I want to touch on just a few ways my written
testimony suggests we can sustain AI success.
First, broader AI innovation and adoption like we are
seeing requires maintaining a dominant research base--abundant
research flows generate talent--but also the ideas,
intellectual property, innovation, and startups that we are
hearing about.
I will provide more detail in the written remarks, but the
Nation remains--does retain clear global leadership. With that
said, it has fallen short of the doubling of AI research that
have been suggested in 2019 by the National Security Commission
on Artificial Intelligence, an important bipartisan watchword.
Beyond that, to address the scale issue, the Nation should
prioritize a step change in total AI R&D outlays, at the same
time seeking to rejuvenate the AI Research Institute's program
for accelerating research on new topics, establish a new
testbed programs involving Federal research units, private
sector, and even data centers that can be brought into this,
and then direct grant making toward investments in what some of
us call pro-worker AI that supports such values as education
systems, health, human learning, and human decisionmaking.
Accelerating AI adoption also requires promoting the growth
of emerging AI clusters in geographic regions. Nvidia, OpenAI,
Microsoft, and Anthropic have all advocated for this kind of
development, sometimes using the term economic zones to
envision regional investment areas that fuse permit and energy
solutions with Federal and state creation, small business
empowerment, and community level prosperity.
Such regional development, including through linkages with
local testbeds, could foment a powerful surge of optimism in
communities. We believe that bottom-up economic innovation in
regions is an important part of this work ahead of us.
More boldly, Congress and the Trump administration could
build new prize competitions or recent challenge grant
programs, such as the tech hubs' and NSF's innovation engines
to further develop these regions and hub engines that have so
much to give.
Finally, AI adoption leadership will further hinge on
ensuring that adequate pools of high-quality talent exist all
across the country and in every sector and every region.
The challenge is both narrow and wide. The narrow part of
the challenge is the Nation's diminishing homegrown share of
the world's elite talent. Speaking of this, ITIF has shown that
while the U.S. attracts and employs a high share of elite
talent, its domestic production of such talents has slowed.
That is a worry and needs to be addressed.
At the same time, concerns around the broad degree of AI
readiness are important, and starting now virtually all workers
will need to understand AI principles, be able to understand
and direct AI effectively, and be able to evaluate its outputs.
AI literacy instruction, as the Department of Labor has
been describing in recent months, will need to cultivate
agility and install such human skills as judgment, teamwork,
creativity, and problem-solving.
If that is widely achieved, workers will feel optimistic
about the technology. If not, I worry they will not, and so a
supportive AI talent strategy must foster AI readiness among
both elite and general populations.
Fortunately, immigrants and U.S. higher ed are proven
talent sources. Thoughtful visa reforms can help the Nation
retain its edge on elite talent, and even more important is the
need to prioritize broad AI literacy, not just at elite
institutions but throughout the region and regional AI----
Senator Budd. Mr. Muro----
Mr. Muro.--the entire workforce system.
Senator Budd. Mr. Muro, if we could.
Mr. Muro. Yes.
[The prepared statement of Mr. Muro follows:]
Prepared Statement of Mark Muro, Senior Fellow, Brookings Institution
Introduction
Chairman Cruz, Ranking Member Cantwell, and distinguished members
of the committee, I want to thank you for the opportunity to testify
today on how we can maximize and sustain the value for human
flourishing of America's extraordinary artificial intelligence (AI)
sector.
My name is Mark Muro and I'm a senior fellow at the Brookings
Institution. Acknowledging that affiliation, I should note here at the
top that these remarks and anything I say today are my personal views
and do not reflect the views of the institution or its other scholars,
employees, officers, or trustees. These are my own thoughts about a
timely topic.
Nearly every week, America's AI innovators are introducing
incredible new tools and solutions that are allowing more and more
people, firms, entrepreneurs, and communities to expand the reach and
achievement of human skills and expertise. This progress has generated
significant cause for optimism. In recent years, most notably, AI's
capacity to drive productivity, advance science, promote health, and
magnify what humans can do has been demonstrated time and again. These
achievements reflect the power of AI's special ability to weave digital
innovations and human skills into a transformative collaboration.
And yet, for all that, significant pessimism has begun to
complicate the moment. Some fear what has been deemed the greatest
automation technology in human history.i Others worry about
how they will weather likely employment disruptions and about the
uncertainty of what skills will serve them in the future. Still others
worry about the impacts of data center developmentii on
local communities, and the potential for highly uneven geographic
build-out of the AI economy.iii
---------------------------------------------------------------------------
\i\ Daron Acemoglu, David Autor, and Simon Johnson, ``Building pro-
worker artificial intelligence.'' Working paper 34854. (Cambridge:
National Bureau of Economic Research. 2026).
\ii\ Daniel Goetzel, Mark Muro, and Shriya Methkupally, ``Turning
the data center boom into long-term local prosperity.'' (Washington:
Brookings Institution, 2026).
\iii\ Mark Muro, Shriya Methkupally, and Molly Kinder, ``The
geography of generative AI's workforce impacts will likely differ from
those of previous technologies.'' (Washington: Brookings Institution,
2025).
---------------------------------------------------------------------------
In view of that, I want to say a few words about areas where
Federal support can help maintain the sector's momentum, reinforce its
value, and promote optimism about its possibilities. Above all we
should ensure that AI is pro-human and pro-community. In this
direction, my fellow panelists have done an excellent job of inspiring
optimism by detailing some of the possibilities. Now I'd like to follow
them by noting a few areas of policy that have helped businesses like
those we've just heard about innovate and that will now be needed to
sustain more of such innovation.
The nation should build a strong AI-support platform.
To speak about what it will take to sustain AI innovation to
support human flourishing, I want to draw on my work at Brookings on
``AI readiness'' to suggest that the Nation needs to build a strong AI-
adoption platform.iv To that end, I would encourage the
committee and Congress coalesce around a core set of AI-adoption
readiness priorities. I'll touch on five areas of needed attention:
---------------------------------------------------------------------------
\iv\ Mark Muro and Shriya Methkupally, ``Mapping the AI economy:
Which regions are ready for the next technology leap.'' (Washington:
Brookings Institution, 2025).
---------------------------------------------------------------------------
Research
First, broader AI innovation and adoption requires maintaining a
dominant AI research base. Abundant research flows generate talent, but
also ideas, intellectual property, innovation, and start-ups. Given
that, there is work to be done.
To be sure, the United States maintains clear leadership in AI R&D
and consistently outperforms other countries on innovation and
investment. In 2024, for example, the U.S. developed 40 AI models while
the second-best performer, China, produced 15. As to private
investment, U.S. companies invested $109.1 billion in AI research in
2024, vastly outpacing China's $9.3 billion.xiii
Yet recent developments suggest these figures won't be enough for
the U.S. to maintain dominance. According to the International Data
Corporation's (IDC) spending report, China is expected to see an 86
percent compound annual growth rate in generative AI investments from
2022 to 2027, meaning the country's generative AI spending could grow
to represent 33 percent of the world's AI investment--up from less than
5 percent in 2022.xiv
Against this backdrop, Federal AI research funding trends are
concerning. To be sure, U.S. government investment in AI research did
increase from $2.98 billion in 2024 to $3.32 billion in
2025.v However, this came at a moment when the authoritative
National Security Commission on Artificial Intelligence (NSCAI) had
advised doubling non-defense AI R&D investments annually to reach $32
billion by 2026.xvi
---------------------------------------------------------------------------
\v\ NITRD, ``Artificial intelligence R&D investments--FY 2019 to FY
2025.''
---------------------------------------------------------------------------
The nation's current mix of research activities also falls short on
several areas of potential opportunity. First, U.S. private sector
investments dwarf government outlays for basic (mostly academic)
research. This may limit creative leaps forward and progress on novel
use cases in underinvested sectors. Dwindling support for the NSF-led
National AI Research Institutes--launched during the first Trump
administration--is a case in point. Another missed opportunity is the
thinness of AI research and computing flows into high-quality but
farther-flung universities. As of now, the Nation's Bay Area Superstars
and Star Hubs account for about 60 percent of the Nation's R&D flows,
leaving institutions in many regions underserved.vi
Likewise, computational support for basic research in academia too
often remains spotty, likely limiting the Nation's innovation
potential.
---------------------------------------------------------------------------
\vi\ Muro and Methkupally, ``Mapping the AI economy.''
---------------------------------------------------------------------------
And so, the Nation should build out its platform for broader AI
adoption by expanding the scale of AI R&D and improving its character.
To address the scale issue, the Nation should prioritize a step change
in total AI R&D outlays in the next decade. Rather than cut research
outlays, Congress needs to ``run faster'' if it wants to outpace China,
because R&D is a critical accelerant.
At the same time, to improve the composition of overall AI
research, the Nation should expand investments in basic R&D research
and in mechanisms for broadening access to essential computational and
data resources. On this front, increased basic research expenditure
into universities appears critical. But so does rejuvenation of the AI
Research Institutes program for accelerating research on new topics in
new universities and locations. Additionally the build out of something
like the National AI Research Resource (NAIRR) pilot program could
provide a mechanism for increasing more and different researchers'
gaining access to high-speed computing resources and datasets,
including in more and different locations.vii Related to all
of this, it would be valuable if some Federal research grant-making was
directed towards investments in ``pro-worker AI'' that supports such
values as education, human learning, and human decision
making.viii
---------------------------------------------------------------------------
\vii\ Mark Muro and Julian Jacobs, ``The case for promoting the
geographic and social diffusion of AI development.'' Washington:
Brookings Institution, 2024.
\viii\ Daron Acemoglu, David Autor, and Simon Johnson, ``Building
pro-worker artificial intelligence.'' Working paper 34854. (Cambridge:
National Bureau of Economic Research. 2026).
---------------------------------------------------------------------------
In a word, an expanded and enhanced research agenda is critical for
creating a national platform capable of bolstering AI innovation,
entrepreneurship, and adoption in the country.
Regional innovation clusters
Accelerating AI adoption also requires promoting the growth of
emerging AI clusters in geographic regions. Dense, vibrant AI clusters
are national assets that contribute to national progress. Neglecting
such clusters is a missed opportunity that leaves innovation and
adoption potential untapped.
With that in mind, the Nation's overall AI platform should promote
broad AI adoption through region-focused industrial development.
Nvidia, OpenAI, and Anthropic have all advocated for this kind of
development, sometimes using the term ``economic zones'' to envision
regional investment areas that fuse permit and energy solutions with
Federal and state job-creation, small-business empowerment, and
community-level prosperity. Such regional development work could foment
a powerful surge of optimism in communities.
Given that, the Federal government should draw on its recent
experience with ``place-based'' industrial investment to accelerate AI
cluster scale-up in promising regions and sectors across the country.
What might this look like? Some of this region-catalyzing work
could leverage the National AI Research Institutes from the previous
Trump administration, as well as the NAIRR program, by orienting their
research and computational supports toward the needs of promising local
clusters. These steps would support early-stage activities in key
clusters. More boldly, Congress and the Trump administration could
revamp recent challenge grant efforts--such as the Commerce
Department's Regional Technology and Innovation Hubs program and the
NSF's Regional Innovation Engines--to focus new hubs and engines
specifically for AI.ix A number of the current hubs and
engines are leveraging AI technologies already. Why not develop new
centers fully focused on compelling AI verticals grounded in dynamic
regional ecosystems?
---------------------------------------------------------------------------
\ix\ Muro and Methkupally, ``Mapping the AI economy.''
---------------------------------------------------------------------------
Taken together, approaches like these will help ensure the national
AI adoption platform promotes new expansion across the Nation's
regions. They will also foster optimism and counter concerns that AI is
somehow an industry working exclusively for others somewhere else.
Talent
AI adoption leadership will further hinge on ensuring that adequate
pools of high-quality AI talent exist all across the economy, available
to every sector and every region. The challenge is both narrow and
wide.
The narrow part of the challenge is the Nation's diminishing home-
grown share of the world's elite AI talent. Speaking to this, the
Information Technology and Innovation Foundation (ITIF) has shown that
while the U.S. attracts and employs a high share of elite talent its
domestic production of such talent has slowed.x That's a
worry.
---------------------------------------------------------------------------
\x\ Trelysa Long, ``AI is powering the U.S. economy, but who's
powering AI?'' (Washington: ITIF, 2025).
---------------------------------------------------------------------------
At the same time, there are also concerns about the broad degree of
AI-readiness needed across the Nation's workforce. Starting now,
virtually all workers will need to understand AI principles; be able to
understand and direct AI effectively; and be able to evaluate AI's
outputs, as notes the Department of Labor's new Artificial Intelligence
Literacy Framework.xi In addition, such ``AI-literacy''
instruction--as continues the DOL--will need to cultivate agility at
scale and instill such ``human'' skills as judgement, teamwork,
creativity, and problem-solving. If that is widely achieved, workers
will feel optimistic and engaged about AI. If not, they won't.
---------------------------------------------------------------------------
\xi\ Employment and Training Administration, ``Training and
Employment Notice No. 07-25.'' Washington, 2026.
---------------------------------------------------------------------------
And so, a supportive Federal policy platform for AI needs to foster
AI readiness among both top scholar echelons and everyone. Fortunately,
U.S. immigrants and higher education stand out as proven sources of
talent. Given that, thoughtful visa reforms will help the Nation retain
its talent lead. But even more important is the need to prioritize
broad AI education and workforce literacy at higher education
institutions and all across the workforce system, not only in the usual
elite locations. Likewise, Congress should support the creation of
regional AI learning networks, with employer-led, cross-sector
partnerships that serve as training and innovation centers for the AI
economy. Aligned to industry demand, all of this training will go a
long way toward ensuring AI unleashes creativity and optimism among
American workers. Connecting all of this to the emergence of AI
industry clusters near universities and community colleges will help
educate, engage, employ, and retain critical technical talent.
Infrastructure
Boosting regional AI adoption will further depend on the delivery
of key infrastructure that is not now solidly in place, but on which
national leadership depends.
On this front, the AI era is elevating the need for large-scale
chip production, vast data and computational resources, new energy
sources, and the build out of huge data centers in accordance with
important electricity, water, and other permitting
issues.xii To be sure, work has begun on some of these
issues, such as through the CHIPS and Science Act's subsidies for
semiconductor plant construction and the launch of the NAIRR pilot for
giving more scientists, innovators, and educators access to the
computing and data resources necessary for game-changing research.
---------------------------------------------------------------------------
\xii\ Muro and Methkupally, ``Mapping the AI economy.''
---------------------------------------------------------------------------
With that said, AI-related infrastructure gaps stand as major
impediments to regional and national scale-up. The demand for computing
resources and energy is projected to challenge available supplies.
Permitting and grid hurdles exacerbate the delivery problem. And to
many communities, data center siting decisions seem secretive and
disruptive--divorced from regional economic planning.xiii
---------------------------------------------------------------------------
\xiii\ Goetzel, Muro, and Methkupally, ``Turning the data center
boom into long-term local prosperity.''
---------------------------------------------------------------------------
In light of these challenges, the Federal government should work to
facilitate timely, carefully planned, and environmentally sound data
center and power development. A portion of that work must clearly
involve policy and regulatory efforts to bring new clean energy
generation sources and grid links online. Some of this will involve
speeding up the complicated federal, state, and local siting and
permitting process for conventional or nuclear power plants, including
by leveraging suitable public lands or replacing coal plants. But it
will also be important to streamline the permitting processes for clean
energy generation and related transmission capacity.
Otherwise, the Federal government should do what it can to
facilitate strategic data center development. With data center
development at times disruptive and localities increasingly wary of it,
Federal and other stakeholders should work with industry to optimize
the process so that it supports AI build-out that maximizes local AI
gains.
Development could be streamlined and rationalized through the
establishment of AI economic zones within states or through the release
of suitable public lands. Likewise, the government could encourage data
center developers to negotiate beneficial partnerships with local
stakeholders, which would complement construction with community AI
development. In this vein, Brookings has suggested how regions might
trade expedited data center regulatory approvals for shared computing
resources, research collaborations, and talent
initiatives.xiv The Federal government could aid in that.
Federal agencies could create and fund an AI tech hubs program, where
data center developers and the Federal government are co-investors in
region-level tech ecosystems along with universities and regional
firms. Alternatively, Congress could approve funding for AI test-bed
collaborations involving the co-location of national labs, data
centers, universities, and startups on Federal land. The Department of
Energy has already moved in this direction with its plan to leverage 16
Federal land parcels for rapid data center construction on sites that
have in-place energy infrastructure and fast-track-
permitting.xv Some of these sites could be managed to create
community economic development as part of the build out.
---------------------------------------------------------------------------
\xiv\ Ibid.
\xv\ Office of Policy--Department of Energy, ``Request for
information on artificial intelligence infrastructure of DOE lands.''
Request for information.
---------------------------------------------------------------------------
Worker security
Finally, any national platform for regional AI scale-up needs to
include strategies to provide basic worker security. Such provisions
are necessary because successful AI adoption will involve both gains
for many workers and dislocation for others. Minimizing worker
dislocation will smooth adoption, keep talent engaged, and maintain
morale.
Recent work from Brookings shows that higher-tech, higher-value,
information-based industries--especially in cities--are likely to see
elevated levels of AI impact.xvi Specifically, Brookings
analysis suggests 30 percent of all workers could see at least 50
percent of their occupation's tasks disrupted by generative AI in the
coming years, with higher ``exposure'' levels for higher-skill computer
and office activities.xvii While some of those impacts will
enhance worker well-being and create new jobs, others could bring about
sudden task shifts, depressed work demand, or even chronic under-or
unemployment. Others may shred long-reliable pathways for worker
mobility.xviii This matters because such disruption could
produce ``adjustment'' challenges for local labor markets, weaken
confidence in the AI revolution, and undermine support for regional AI
scale-up.
---------------------------------------------------------------------------
\xvi\ Mark Muro, Shriya Methkupally, and Molly Kinder, ``The
geography of generative AI's workforce impacts will likely differ from
those of previous technologies.''
\xvii\ Mark Muro, Shriya Methkupally, and Molly Kinder, ``The
geography of generative AI's workforce impacts will likely differ from
those of previous technologies.''
\xviii\ Forthcoming research from the Brookings Institution and
Opportunity@Work.
---------------------------------------------------------------------------
Given that, the Nation's Federal AI platform needs to provide
elements of a worker-adjustment strategy that helps regions deliver on
Vice President JD Vance's promise that AI adoption will bring workers
``higher wages, better benefits, and safer and more prosperous
communities.'' xix
---------------------------------------------------------------------------
\xix\ Reuters., ``Quotes from U.S. Vice President JD Vance's AI
speech in Paris.'' February 11, 2025.
---------------------------------------------------------------------------
Much remains to be worked out on how to deliver this. But for sure,
the Nation will want to invest more in ``active labor market policies''
that help people shift into new jobs. These policies may involve rapid
retraining programs for individuals impacted by AI-related job
displacement, such as the pilot efforts the Trump administration has
advanced.xx Relatedly, these policies may entail flexible
benefits, including for wage insurance, and health care that is not
tied to one employer. Other supports may include policies that give a
measure of economic security to workers who want to be retrained and
learn new careers. For example, Brookings has described the idea of a
``Universal Basic Adjustment Benefit'' that would help displaced
workers transition to new work with the help of temporary income
support that allows for intensified training access.xxi Such
provisions can provide a measure of stability as the nature of work
evolves while also kindling optimism among workers who may currently be
discouraged due to their fears of displacement costs.
---------------------------------------------------------------------------
\xx\ U.S. Departments of Labor, Commerce, and Education.
``America's talent strategy: Building the workforce for the golden
age.'' (Washington, 2025).
\xxi\ Mark Muro and Joseph Parilla, ``Maladjusted: It's time to
reimagine economic `adjustment'' programs.'' (Washington, Brookings
Institution, 2017).
---------------------------------------------------------------------------
In sum, innovative firms of all kinds--whether AI developers
themselves, the increasing millions of AI-adopting firms, or the
thousands of AI start ups entering the space--are providing abundant
grounds for excitement about AI's future potential. Their
entrepreneurship gives much cause for optimism. And yet, sustaining
that optimism requires sustaining those firms' growth and Americans'
confidence in the future--and that means reinvesting in the
fundamentals of American AI strength and vibrancy.
Which is why Congress should build a strong AI-support platform,
one that begins with robust investments in R&D and regional innovation
clusters, and that leans in on talent, infrastructure, and worker
security. Implemented well, such a platform will foster both continued
growth and a broader, more widely shared confidence in the AI future.
Senator Budd. Thank you so much. If we could take a few
questions for the rest of the panel.
Mr. Muro. Absolutely.
Senator Budd. If we have additional thoughts, we will come
back to that and--when you are recognized, if that is OK.
Mr. Muro. Great, thank you.
Senator Budd. Yes, thanks again for being here. Thanks for
your opening remarks.
You know, it has been said that the four main inputs of AI
are talent, compute, energy, and data. The Federal Government
houses a tremendous amount of scientific and important datasets
that could be leveraged as strategic assets in the U.S. AI
leadership.
The Open Government Data Act of 2018 requires data assets
owned by the Federal Government whose sharing would not
otherwise be prohibited by law to be published in machine-
readable format.
If each of you first three with particular companies--
operating companies--what would each of your companies' efforts
to deploy AI--how would that be affected by more access to
data? If it is AI-ready and if it is machine-readable and
compatible, how would that affect each of your companies?
We will start with you.
Mr. Giannikopoulos. Senator Budd, the access to data is a
critical aspect of development for artificial intelligence.
However, in health care the ability to have validation data
by which you can measure the quality of your solutions that are
developed by which deployers at the institutional level can
assess the fit within their personal institution is one of the
greater challenges.
So having more robust access to this machine-readable,
translatable, and ideally in health care outcomes-linked data
will provide opportunities to actually assess these solutions,
not in a vacuum but against real American data as part of that.
So, you know, more robust access to that would be a
significant enabler of adoption and innovation of this new
technology.
Senator Budd. So in the world of radiology, what would be a
specific dataset that you would look to make more accessible?
Mr. Giannikopoulos. Radiology, there are two key areas. It
is the images and the reports, and the reports in radiology are
one piece of siloed information. Yet, what happens with that?
If a--that quiet lung mass that I referenced in my opening
statement, if that is identified what percentage of that turns
into cancer? How do you understand the progression of that?
Being able to access that longitudinal information at scale
would enable significant development.
Senator Budd. And for something like that in particular
with radiology, that would not violate patient privacy?
Mr. Giannikopoulos. If appropriately de-identified, which
would be a very important aspect of this, absolutely not.
Senator Budd. Great. Thank you for that.
Ms. Ng, what are your thoughts for Siemens Industries?
Ms. Ng. Chairman, thank you for the question.
More available, highly trusted data is gold in industrial
AI environments and something that is very important to
recognize is that for industrial AI, the underlying data comes
from machines. It comes from manufacturing systems and
engineering datasets within the four walls of that factory or
the shipyard or production environment.
So the more readily available that data is and in usable
formats, the better the generative insights and recommendations
that the industrial AI products can make.
Senator Budd. That is very helpful, thank you.
Dr. Shelton?
Mr. Shelton. So I think from an operating company
perspective, we have to remember that there are really only
three ways that we can actually have access to data.
The most expensive is we can go out and generate it
ourselves, and you see robotics companies trying to do that
today by a tele-operation.
The second is when the data is available on the public
Internet for free and this is one of the ways that the modern
AI frontier labs have made such rapid progress training their
LLMs specifically.
But that data does not exist for robots generally. There is
no description of human movement or industrial tasks that you
can really find on the open internet.
And then the third way is simulation, and although there
has been a lot of rapid progress--I am sure you have seen news
reports with Nvidia and others who have done that kind of
work--we are still in the early days of that really being
viable for large scale training of robots.
So the government release of data is actually helpful a
couple of ways. First is it gives us a very broad look at the
economy as a whole.
Although Agility has been focused to date on these internal
warehouse jobs, there are broader classes of job that would
pull on things like large geospatial datasets and other large
public data sources that would be very exciting to have access
to.
And second is I think it gives us a broad representative
look at what people are doing. That is part of the reason why I
cited the census data is that the government has a very large
role to play here identifying areas where there are
opportunities for robots to get out and do useful work. So I
will excited to be a part of that.
Senator Budd. Dr. Shelton, do you have an example of those
ideas inside the census data?
Mr. Shelton. Yes. I mean, I think what is fascinating with
this is how large of an opportunity there really was with
underemployment in sectors like the warehouse and logistics
world.
Feeding things like accident data, human performance data,
and stuff allows us to focus on the very high-value tasks that
actually improve worker safety, improve worker health, and
allow those workers to redeploy elsewhere within the org.
So we are excited about being able to tailor what our
product offerings are to the needs of the broader workforce.
Senator Budd. Dr. Shelton, do you think that would help
with adoption? I mean, there is a general fear out there about
job replacement as sort of a generalized concern.
But if you show massive areas of underemployment in a
sector such as warehouse and logistics where there is need for
much more workers, do you think that would help with adoption
in those areas?
Mr. Shelton. Absolutely. In fact, we see a lot of pressure
from the small and medium business community as well that would
love to adopt this kind of technology and, you know, it is--
without name checking a specific retailer, we were approached
by a small but nationwide retailer a number of years ago,
looking for a back warehouse automation solution because they
were having staffing challenges.
Unfortunately, it was not possible at the time with the
technology that was available to us circa 2020 to really engage
with them. But I think the broader look at the labor pressures
that are faced by those businesses in particular would be super
helpful.
Senator Budd. To give us a quick snapshot of the timeline,
if you had that conversation in 2020.
Mr. Shelton. Correct.
Senator Budd. Now in 2026, if you had the same
conversation, do you have more capabilities?
Mr. Shelton. We do. I think there is a rollout question
that we have, which is do you tackle the low hanging fruit
first or do you go for the harder end of the spectrum.
But, certainly, technologically the advances that AI has
shown just in the last I would call it 3 years significantly
lowers the barrier to entry to address those businesses.
Senator Budd. Thank you.
Ranking Member.
Senator Baldwin. So the administration suggests through its
recently released Maritime Action Plan that AI can be used to
increase domestic shipbuilding capacity. Currently, the United
States produces less than one percent of ships globally.
We used to be dominant in the industry. The effort to
revitalize domestic shipbuilding requires cooperation and input
from government, the private sector, labor, and more.
So, Ms. Ng, where do you see the most promising
opportunities to use AI to increase domestic shipbuilding
capacity and how is Siemens working with organized labor and
workers in the shipbuilding sector to identify opportunities to
use AI to expand and enable our workforce to increase output,
including repair and maintenance?
I am going to give you a cluster of questions here. Do you
think we can attract workers to expand our capabilities if they
continue to hear that AI might displace their jobs?
So go at it.
Ms. Ng. Ranking Member, thank you very much for the
question.
I will start by sharing a little bit about what industrial
AI can do to recapture the American dominance of shipbuilding
here in the U.S.
What we are seeing is that by the usage of industrial AI in
all aspects of the life cycle for shipbuilding is really
critical, and I will give three quick examples.
First of all, in design. Industrial AI allows us to be able
to test and model thousands of configurations of a ship in
hours as opposed to months. This allows us to do quicker design
churn and to be able to design ships very quickly.
Second, we are seeing with production we are using
industrial AI to help basically simulate material flow,
workforce allocation, and sequencing of work packages so that
work done in a back shop, on a shipyard, is very efficiently
sequenced and done quickly.
This allows us to execute on that work and on those great
designs even faster.
The third example that I can give is in sustainment. We are
using industrial AI applications with predictive maintenance to
be able to preserve uptime so that all of our machines and all
of our capabilities in the shipyards and in the back shops are
able to operate efficiently.
We are seeing that bringing that entire life cycle view
together, we are very able to be able to reestablish that
dominance.
The other question that you had around how do we continue
to support the workforce and make sure that we are being able
to retain that workforce is, I would say, pretty
straightforward.
We have to have a strategy to attract workforce and retain
them, meaning the workforce is becoming increasingly digitally
native. They have expectations. They want to work with top-of-
the-line software and AI-enabled tools.
So we are really honored to be able to provide that as part
of talent acquisition strategies for manufacturers and
shipbuilders specifically.
Senator Baldwin. And I certainly want to just encourage you
to have them at the table from the beginning, not introduce
them at the end.
Ms. Ng. Absolutely.
Senator Baldwin. Mr. Muro, I have a question for you.
Last month, you co-authored a piece titled ``Turning the
Data Center Boom into Long-Term Local Prosperity.'' In this
article, you note that local officials have the leverage to
engage data center developers on becoming true local partners
in their community.
The article highlights some efforts in Wisconsin,
specifically Microsoft's partnership with the University of
Wisconsin-Madison, in the development of the AI Co-Innovation
Lab and their partnership with Gateway Technical College to
train workers.
What should local officials be trying to get out of these
negotiations? We have got a lot of them going on in Wisconsin.
Mr. Muro. Wisconsin actually is a demonstration, almost
like an aerial view of data center development that is
benefiting a region that is tied to broader economic
achievement for the region, too.
That is because the--they are technology themselves. They
are a source of computer processing, high-speed technology of--
that could be tied in and is being tied into regional
university activity, research, and so on. And then there are
users of energy and places for energy innovation.
So data centers can be thought of in a very broad way along
with their important national demonstration or their national
push toward powering the whole technology.
But places have the possibility to work--enter into
interactions with the hyperscalers early in the process. We
think there is an opportunity for them to trade essentially
very quick initial ability to save--excuse me, to establish,
you know, quick permitting in response and in a trade with the
companies to build these kind of partnerships in the region.
So I think the more a region has a sense of its technology
goals and where those intersect with the hyperscalers, and I
think there are lots of areas for particular research, work on
energy issues, testbeds of all sorts. So I think there is just
a wider array of possibilities in regions.
Senator Budd. Thank you, Mr. Muro.
Senator Cruz, you are recognized.
STATEMENT OF HON. TED CRUZ,
U.S. SENATOR FROM TEXAS
The Chairman. Thank you, Mr. Chairman. Welcome to all the
witnesses.
Let us start with a broad question. AI is a fundamentally
transformational technology, and like past waves of innovation,
we do not know exactly how it is going to impact our economy,
including what it is going to do to employment.
One of the greatest fears that I hear from people is
concern that AI is going to take their jobs and it is going to
lead to fewer jobs.
Mr. Giannikopoulos, what do you say to those concerns? What
are you experiencing in your current job markets and how do you
anticipate AI shifting the job market in respective industries?
Mr. Giannikopoulos. Chairman Cruz, thank you for the
question.
The crisis in health care is real and it is happening right
now. There are around 900,000 physicians in the United States.
The projected shortage of physicians by 2030 is 187,000--excuse
me, by 2037.
The projected shortage of nurses by 2030 is 194,000. So the
gap is growing. The need is there, specifically within
radiology. If you look at the attrition rate over the past 5
years, it is up by 50 percent compared to historical norms.
Meanwhile, the projection is that imaging volume will rise
by 26 percent in the next 30 years. So a continued growth and a
continued need.
Dr. Curtis Langlotz, the immediate past president of the
Radiological Society of North America, recently did a task-
based analysis of the work of a radiologist, and he estimated
that within the next 5 years the amount of work a radiologist
will need to do could possibly go down by 33 percent.
Yet, the need for them to perform that understanding and
ultimately, you know, make those diagnoses so the patient can
get the care will only increase.
So AI is not replacing the physician at any point along
this. It is enabling them so that we can leverage them to get
where we need to be as a health care society.
The Chairman. Ms. Ng, how would you answer the same
question?
Ms. Ng. Chairman Cruz, in our experience, advanced
technologies support workforce expansion, not contraction. This
is especially true in my sector of shipbuilding, where we are
seeing these advanced technologies drive an increased demand
for highly skilled trades.
AI is not diminishing the need for skilled workers. It is
amplifying it. AI's expansion reinforces the need for strong
technical skills and for people who are passionate about
innovating it.
The best example that I have is actually in Fort Worth,
Texas, where we have the latest and greatest Siemens facility
that was just stood up. In Fort Worth, we used our digital twin
and AI-enabled tools to be able to model the entire production
floor and simulate it before even breaking ground.
This facility created hundreds of new jobs in that area,
and was--and the employees that are working there are operating
in reduced complexity and much better training.
So in summary, the greater risk to jobs, in our opinion, is
that--is losing industrial competitiveness. It is not
responsible modernization. Thank you.
The Chairman. Dr. Shelton.
Mr. Shelton. Yes, thanks for the question.
I can give two examples with human labor. The first in the
warehouse and logistics world is there are sort of two
problems.
First right now is an extreme growth in that sector over
the last, say, 10 or 15 years--all of us love our next-day
Prime shipping--has created a completely unsustainable growth
trajectory in that industry.
So there are only two ways you can do that. One is you
could pull slack out of the rest of the economy from a human
labor standpoint, and second is you can deploy automation.
Companies right now have tried to do both, like the Amazon
example that I cited. However, as a sector, it is still coming
up short from a total employment count standpoint. So that is
the backfill side of this.
The second and the more exciting piece of this is I think
it changes fundamentally the nature of what a job is. So as a
private pilot, there was a time in the U.S. history where
delivering air mail meant you put on your goggles, you got into
your biplane, and you flew across the U.S. personally carrying
the mail.
There is now an enormously complex logistics system for
shipping things by air, and you could say that you work in that
industry while someone from the 1930s would simply not
recognize what the job has evolved into.
That is why I got into robotics. That kind of stuff is
super exciting to me and the evolution of things like
agriculture. As a notoriously terrible grower of corn, I think
people do not--are not aware of how hard modern agricultural
jobs are and just how far we have come on that side of the
economy.
So I am an optimist about this. I do think we should be
cognizant that there are evolutions of jobs over time and we
should be sensitive to that, but it is also the path forward.
The Chairman. Two final questions. What is something that
you hope AI could do in your industry that we cannot currently
do yet, number one? And number two, what is a surprising way
you have seen AI used that you did not expect?
Mr. Giannikopoulos. I will answer the same way, but from
two different perspectives. If you look at precision medicine
pathways, AI has already opened those up in ways that we have
never seen.
Take my personal diagnosis journey. It was 10 years to
that, ultimately--10 years of varying symptoms that were
ultimately dismissed because I was a white male with Greek
heritage in Florida, which is not exactly common for multiple
sclerosis.
Being able to identify that this, you know, varying
numbness, all these different parts and pieces personalized to
me as an individual and understanding in our broader health
care system, could have shortened that cycle of diagnosis.
That is what it is already starting to do, but with better
integration, better understanding, better tailoring of the
medicine and the understanding to the individual, it will be
able to take that to the next level.
This is where we are getting into, you know, predictive
cancer scores, different things like that, so that you can
identify a path that a person is on, not when they are already
on it but before they start it, and really make a big
difference.
The Chairman. OK. Briefly, Ms. Ng, and then Dr. Shelton.
Ms. Ng. To answer your first question, we want to see AI
adopted throughout the entire life cycle as opposed to in silos
so, say, maybe just design or just production or just
sustainment.
For the second question, what has surprised me the most is
seeing the application of industrial AI with the U.S. Navy,
which is my customer.
We have been able to work with the Navy to digitally model
the four shipyards and help them to be able to plan for future
work.
So when you build a submarine, you have a whole maintenance
plan that comes associated with that, and what is surprising to
me is the ability to use industrial AI to actually model out
new dry docks, new back shops, all of this new physical
infrastructure that does not disrupt the current availabilities
and work happening there.
Thank you.
Mr. Shelton. So the very first time we deployed one of our
robots, Digit, the humanoid, doing a task, it took an engineer,
as I recall, somewhere about five and a half weeks to really
prototype that.
By 2023--and this video is actually on our YouTube channel
for anybody who is curious, we decided to hook an early version
of ChatGPT up to it to see if it could write code, if it could
run the robot.
And shockingly, and I was completely gobsmacked by this, it
worked on the first try. Nothing in technology ever works on
the first try so that was a really interesting outcome.
Now, at the time, it was not hardened. It was not
deployable. But as sort of a shot across the bow of things that
were to come, it was super interesting.
I would love--and this gets back to Chairman Budd's
question about the dataset availability--that that worked is a
sign that we can take descriptions of tasks that we want robots
to do and use automated tool sets to get them deployed as
rapidly as possible.
It is an exciting and completely novel side of the industry
that continues to surprise most of us who have been in the
field for a while.
The Chairman. Thank you.
Senator Budd. Thank you, Chairman.
Senator Blunt Rochester.
STATEMENT OF HON. LISA BLUNT ROCHESTER,
U.S. SENATOR FROM DELAWARE
Senator Blunt Rochester. Thank you, Chairman Budd and
Ranking Member Baldwin, and thank you so much to the witnesses
for this hearing.
I get very excited about this topic. I was former secretary
of labor in Delaware as well as head of state personnel.
But I also had the opportunity in the House when I served
in the House to start a bipartisan Future of Work Caucus,
because for me, if anybody says they know what the answers are
or whether the economy is going to--you know, we are going to
have more jobs or less jobs, you really do not know.
This is all so new and so--and so impactful in so many
parts of our lives. I am on the HELP Committee, so health care.
I have a nursing workforce shortage bill, so health care is
important.
I am from an ag state so we think about precision
agriculture. I saw a woman use an iPad in her kitchen to
control her crops.
So to me, this is very important. Our state has a statewide
AI commission and we are setting up clear guardrails on ethics,
safety, transparency across models, on training, creating a
regulatory sandbox, and also new legislation.
We also have two land grant institutions, the University of
Delaware and Delaware State University, that are actually
turning AI into real-world innovation and workforce pathways,
and we have a lot of local companies like Qnity who are
deploying new hardware to improve AI systems.
And we also have to pair AI adoption with real safety
guardrails and policies that ensure that we strengthen our
workforce instead of sidelining it, and I think, you know, that
is one of the--for me, there is the--I am a pragmatic optimist.
Let us put it that way. So there are the good things, but I
also understand that we do not want to leave people behind.
Whether it is a podcast, whether it is magazines like the
Atlantic, the AI and the Future of Work, it is truly
everywhere. We are reshaping how we work.
I want to really kind of tailor my questions to something,
kind of picking up on Senator Baldwin about the workforce and
how the workforce is included as these changes are happening,
and I will start with Ms. Ng.
What does engagement with your workforce look like when it
comes to designing and implementing AI-related changes? Can I
start with you?
Ms. Ng. Absolutely. Thank you for the question.
As I mentioned in the opening statement, what we have found
at Siemens is that digital transformation truly does work best
when the worker is at the center, and that is over a multitude
of ways.
The best way that I can provide a story, though, is
actually a story from the Puget Sound Naval Shipyard. There is
a superintendent of Backshop 31 at Puget Sound and she truly
demonstrates what effective partnership looks like when we are
trying to solve a problem using a digital technology.
And what I mean by that is this superintendent, she leads
from a place of empathy. She knows the struggles and the pain
points that her workforce is managing. So working with leaders
like that.
Senator Blunt Rochester. Do you have like a--is there a
specific--and for all of you, I will just submit questions for
the record because I have a lot of questions and I do not have
a lot of time.
But I will follow up with you on, like, is there a specific
process that you use of a way to get that information and then
turn it into your policies?
I know that, Mr. Muro, you said in your testimony Brookings
analysis suggests 30 percent of all workers could see at least
50 percent of their occupation tasks disrupted by generative AI
in the coming years.
So even with the updating of tasks, can someone share with
me how you include workers in making sure that their jobs--they
are re-skilled, that they are trained, that they have what they
need to keep up?
And I will start with--maybe I will start with you, Mr.
Muro. And I have 41 seconds.
Mr. Muro. Yes. No, it is--first, those statistics are
directional. Clearly, there is tremendous uncertainty around
the--but we are talking about a pathway. We know that there
will be significant disruption.
Disruption, though, is--can be positive or negative and we
think significant aspects of AI's impact on work will be
disruptive, but also beneficial.
Senator Blunt Rochester. I mean, I understand the
difference between a horse and a buggy, and now we have cars. I
get that part.
I am looking for--and we will follow up with each of you--
what specific things are we doing to prepare the workforce and
also include them in the help and decision-making of how to
make sure that it is not so disruptive?
I love the issue of coding. We were pushing people to go
into coding, and now the machines can code.
And so as I am thinking about the future and how I think
Dr. Shelton said the changing nature of the workforce, I would
love to have a conversation with each of you about what can we
do as Congress to ensure that we are not leaving anybody
behind, but at the same time we are benefiting from the
technology that is before us.
And I am out of time. I yield back.
Thank you, Mr. Chairman.
Senator Budd. Thank you, Senator.
Senator Blackburn.
STATEMENT OF HON. MARSHA BLACKBURN,
U.S. SENATOR FROM TENNESSEE
Senator Blackburn. Thank you, Mr. Chairman, and to each of
you, thank you for being here.
Establishing a framework for AI is something that is going
to be very important for Congress to do. President Trump asked
me to take the first stab at drafting the Trump America AI Act,
and it is based off of his executive order.
I was pleased that the executive order included protections
for what I call the four Cs: children, creators, communities
from job loss and high electric rates, and then censorship.
We know that AI and the LLMs are biased against a lot of
conservatives, and in Tennessee I like to say we have a good,
bad, and ugly relationship with AI. Our manufacturing,
logistics, health care, really like it.
A lot of our innovators that hold patents and trademarks
and work with our auto industry and then, of course, our
entertainment industry with our musicians, our songwriters, our
screenwriters, our script writers are really quite concerned
about what it is going to do and their ability to protect their
name, image, likeness.
So having the proper guardrails in place are important, and
we know that every industrial sector has had those guardrails
except the virtual space, and once we define what the rules of
the road are, we free up innovators to take off and do great
work because they know what the playing field is so thereby
they can develop a strategy to win.
So that is the purpose of our having the Trump America AI
Act and putting these protections in place, and I want to--I
think it is Mr. Giannikopoulos. Am I saying that anywhere near
right?
Mr. Giannikopoulos. Very close, Giannikopoulos.
Senator Blackburn. Kopoulos. I did get very close.
Mr. Giannikopoulos. You did.
Senator Blackburn. All right.
As you know, Nashville area really is the health care
informatics and health care interactive technology hub, and we
are seeing so much innovation that is taking place there and we
are really--when it comes to predictive diagnosis, disease
analysis, remote surgeries, telehealth even, we see the benefit
of these applications.
The but comes into place when we talk about creating a
digital health ecosystem and privacy concerns that are there.
So in your work, I want you to talk about that importance as an
innovator and a patient.
I am going to give you a total of a minute, and then if you
want to submit something longer form to me, I would welcome
that.
Mr. Giannikopoulos. Excellent. As a developer within health
care governed by business associate agreements, we are, you
know, clearly within the scope of HIPAA. HIPAA is an elastic
rule. It started out as an insurance----
Senator Blackburn. It needs to be modernized.
Mr. Giannikopoulos. It does need to be modernized, but it
started out as an insurance portability.
Senator Blackburn. In order to include all of this.
Yes, as I say, it came about--it covers the fax and the fax
machine, the paper in the fax machine. So we can modernize it.
Mr. Giannikopoulos. It covers the access to data and with
electronics we have the ability to audit it.
Senator Blackburn. That is right.
Mr. Giannikopoulos. That is really key with the
transformation we have gone through.
Nashville, as you have mentioned, has led the pack in that,
and what I see out of the large institutions is they are
approaching this as workforce transformation. It is not simply
leveraging technology to do what you have always done.
It is how can we leverage the technology to do something
new with it, and as a patient that is where the benefit sits
for us, if we can receive better care pathways as a result of
that.
Senator Blackburn. Ms. Ng, I want to come to you and your
work.
Let us talk about the difference between industrial
applications and commercial applications, because I think this
gets lost many times as we talk about how consumers approach
the utilization of a technology and their expectation of that.
And as we have developed tech policy through the years, I
have told people, look at the computer and look at the backside
and the front side, and there is truly a difference that is
there.
And with AI and the fact that AI, coupled with the other
computational sciences like quantum, is going to yield faster,
more accurate results. But talk about the difference in the
industrial and the commercial.
Ms. Ng. Thank you, Senator.
You touched on it perfectly, which is that in a
manufacturing and industrial environment, oftentimes the work
is safety critical and operating in a high tolerance
environment, meaning that precision matters extremely
importantly.
So consumer AI is using, you know, all of the Internet to
analyze data and do various things. Industrial AI is using much
more controlled datasets from machines on the floor, from
manufacturing systems, and from digital engineering datasets,
and it is using that control data to generate insights and make
recommendations.
Ultimately, the human is still the decision authority to
confirm what to do next.
Thank you.
Senator Blackburn. And, you know, as we look at training
our children and we need to have STEM labs in all of our
schools. Even though much coding will be done by AI from the
consumer side, kids need to code just like they need to learn a
language. They need to have that understanding.
So we look forward to helping build this policy out as we
move forward. Thank you.
Thank you, Mr. Chairman.
Senator Budd. Thank you, Senator Blackburn.
Senator Moreno.
STATEMENT OF HON. BERNIE MORENO,
U.S. SENATOR FROM OHIO
Senator Moreno. Thank you. I did not see where--I did not
see that pattern coming. That is great.
Thank you for this hearing. It is, obviously, very, very
important and very topical.
I will start with you, Demetri, because I am not going to--
I am not going to take a swing at your last name so I will just
call you Demetri. We are going to make it informal, part of I
will call you by your first name, so you can call me Bernie.
In the 1920s, would it surprise you that there was people
saying machines will devour man?
Mr. Giannikopoulos. I believe they were called cars.
Senator Moreno. Exactly. And in the 1950s, an MIT professor
did a study that said that basically we are stumbling into the
automation era and all workers will be replaced. Does that
sound familiar as well?
Mr. Giannikopoulos. Absolutely.
Senator Moreno. And in the 1980s, there was a book written,
``The Robot Is After Your Job.'' Does that sound familiar as
well?
Mr. Giannikopoulos. The Excel spreadsheet was also going to
replace accountants.
Senator Moreno. Right, exactly. So the doom and gloom has
been around forever, and I think it is important for the record
is just to talk about employment numbers in America from the
1920s to the 1950s to the 1980s.
So in 1920s, when machines were going to devour man, there
was 42 million Americans employed. In the 1950s, that number
went to 63 million. So, obviously, decently wrong.
In the 1980s, that number is 108 million, and in 1990s with
the Internet that was going to replace all retail businesses.
There was no point in being in retail--I remember that being in
the car business--because the Internet was just going to
completely disintermediate everybody, 131 million people
employed.
In the 2000s, at the beginning of smart phones, 160 million
people employed, and thanks to the policies that we put in
place in the--this Congress, we now have a record number of
Americans employed, 172 million.
So the Mark Twain expression, ``The news of our death is
sorely exaggerated,'' meaning employment continues to grow.
Now, with that said, I think what we can all agree on--and
I want to get your thought, and I will start with you,
Brittany, sorry, because there is no way I am taking a shot at
your name either--on what is different about this period of
time, from your point of view, with the introduction of AI
versus what I have just laid out?
Ms. Ng. Thank you for the question, Senator.
There is an incredible story about Albert Einstein riding
on a train, you know, back in the day, and the conductor is
coming through, stamping tickets, and he sees Einstein
scrambling, looking for his ticket.
And he looks up and says, ``Mr. Einstein, it is OK, I trust
that you bought a ticket. I know who you are.'' Continues on
his way. Sees Einstein still scrambling on the floor.
Goes back and says, ``Mr. Einstein, truly, it is fine. I
know who you are. I know that you bought a ticket.'' And Albert
Einstein looks at him and says, ``Thank you, sir, I also know
who I am. What I do not know is where I am going.''
And I think that that story really gets at the heart of
what you are mentioning, which is ultimately what we need to be
doing now is partnering with industry to define what that
compelling vision and that roadmap is, and that is what I do
every single day on the ground at Siemens with my customers is
partnering with them to define what are the actual business
outcomes that you are trying to achieve, and how can we fully
channel these industrial AI solutions and applications toward
achieving that outcome.
So to answer your question, I truly believe that what is
interesting and new about this technology is that we have the
ability to explicitly help specific business outcomes.
Thank you.
Senator Moreno. Yes, and obviously, productivity
improvements is key, and I think the expression has been used,
you are not going to lose your job to AI but you may lose your
job to somebody who uses AI.
And, Dr. Shelton, I will kind of shift to you on this part
of the questioning. Part of what I see as a worry is that this
is moving much faster than these other rollouts, right?
So this is definitely at a speed which we are not used to,
and what I worry about is are our institutions that train
people, are they prepared for the speed?
What are your thoughts on that?
Mr. Shelton. It is a qualified yes, I think, to your
question, although I would agree that the speed here is very
different.
I think something that everybody in this room has had to
grapple with with AI is we did not perceive automation coming
for our jobs or evolving the nature of work for us as, largely,
knowledge workers.
And I think, you know, ironically, people more on the
manual labor and the blue collar side of the universe has had,
you know, several hundred years to get used to the idea of
using tools to accomplish more than they could have
accomplished on their own.
That is why I used the farming example in my opening
statement. Those of us who deal primarily on the knowledge side
of the universe have not had options to deploy autonomous
agents until the last, let us call it 6 months, to go off and
do portions of the tasks that we are using them for, not to do
the job, per se, but to do a sub-component of it.
So I think all of us who are on the more knowledge end of
the economy are going to have to evolve very rapidly to start
understanding our job as managing fleets of autonomous workers
who are working on our behalf as opposed to doing that work
ourselves, and that is quite different.
From a university standpoint, I think there is going to be
a rapid pivot back to the liberal arts. This is something that
my wife, who is an English professor, likes to remind me with
some rapidity is she has known all along that it was all about
talking, and I have to concede that that is probably looking
like it is going to be the case over the next 20 years.
Senator Moreno. All right. Thank you.
Senator Budd. Thank you, Senator Moreno.
Senator Hickenlooper, you are recognized.
STATEMENT OF HON. JOHN HICKENLOOPER,
U.S. SENATOR FROM COLORADO
Senator Hickenlooper. Thank you, Mr. Chair. Thank you all
for being here. Certainly, a very timely hearing, and I commend
the Chair for that.
Let me start just to talk a little bit about product
testing.
Mr. Giannikopoulos--I think that is close--thank you for
sharing how Rad AI is integrating the software to support
quality--the delivery of quality health care to patients.
As you know, software, hardware, are rigorously tested
before they are used with patients. There is endless testing
and backing up.
We are working on bipartisan legislation called the Vet AI
Act, which promotes evidence-based best practices to help
companies have their AI tools independently evaluated by
independent third parties. Let us leave it at that.
This helps increase transparency, promote accountability
for responsible system design. I think, to a large extent, it
creates trust in customers.
So, Mr. Giannikopoulos, would you describe how--what Rad AI
is doing now to test--you know, to looking at what are its--
whether its products are sufficiently tested before they go
into action?
Mr. Giannikopoulos. We are a company founded by a
radiologist for radiologists. Dr. Jeff Chang, the youngest
radiologist at the time of his graduation, 17 years of age----
Senator Hickenlooper. Jesus.
Mr. Giannikopoulos. Yes, a very, very smart man. Has been
central to the design ethos of our solutions. He was a
practicing radiologist when he created the idea of Rad AI to
begin with.
We have continued to engage with the radiologists, validate
the usage, and understand exactly how it integrates into the
workflow.
The other key piece that really needs to be built with the
adoption of AI in health care in particular is trust, and a way
to gain trust is--there is the technical nature, there is the
transparency, the information about how you have built your
models.
But there are also leveraging relationships. For example,
we have one with the RS&A Ventures Group to integrate the
century of knowledge that RS&A has available directly into the
application at the point of care for the radiologist.
So it is not AI assisting them with the diagnosis, it is AI
plus trusted and validated information right at the point of
care.
Senator Hickenlooper. Cool. It is worthy of a longer
discussion.
Ms. Ng--is that right? Somehow close? See, I am not scared
of these, well, you know, difficult names. Unusual names, we
will call them. I have a difficult name, Hickenlooper.
Thanks for your testimony and sharing how all the work that
Siemens is doing around, you know, how AI is going to enter new
sectors of the economy and especially the industrial economy.
It is clear that we are in the middle of a revolution. I
think of it as a great transition and I think we are doing
several things at once. We are going into AI. We are moving
toward clean energy. This will look back--50 years or 100 years
from now, they will look back into the beginning of this great
transition.
I think it is going to--in the terms of AI, it is going to
define how AI supports this growth of small businesses and
entrepreneurs, but also attracts students to STEM fields and
ultimately will transform our workforce.
Now, when two companies enter into a contract, it is
essential that the service agreements are transparent but also
that they are enforceable for--well, for how an AI system can
be used or not used or misused.
So, Ms. Ng, how does Siemens ensure it is transparent with
its customers about the design limitations of the AI products
it sells? Do you understand what I am asking?
Ms. Ng. Yes. Thank you, Senator, for the question.
So transparency is critical to trust, which we were just
speaking about, and there are several ways that we as Siemens
ensure that we are providing full visibility into not only how
industrial AI is being used in environments like a beer
manufacturing customer that we have in Colorado, but throughout
all of the sectors in the United States.
And it really comes down to two things. One is allowing the
industrial AI application to be able to share the source data,
so pulling from which machines, which engineering datasets, and
which manufacturing systems that it is producing from.
The second is providing explanation or evaluation of why
that recommendation is being made, which comes down to the
ability that we were speaking about earlier of training folks
to be able to work in systems oversight types of roles so that
they can engage more effectively with that industrial AI
capability.
Thank you.
Senator Hickenlooper. Great. I am going to hold off. I have
a couple other questions but I will submit them in writing. I
appreciate--they are slave drivers here in terms of keeping us
on our time.
Senator Budd. Only because more showed up. So thank you for
the question.
Senator Young, you are recognized.
STATEMENT OF HON. TODD YOUNG,
U.S. SENATOR FROM INDIANA
Senator Young. Well, thank you to our witnesses for being
here today.
And Ms. Ng, I would like to build maybe on the previous
questioning and your response. We need to unlock more
industrial data, spatial data, as I understand it, a lot of it,
so that embodied AI or AI used to control machines in the
physical space can continue to move forward.
China, a country we care about because we compete with them
on a number of different levels, they lead the world in terms
of deploying industrial robots, collecting this sort of data.
We are going to have to come up, of course, with a different
set of rules and practices, ones consistent with our values and
laws.
But I think for starters, we need to distinguish between
industrial AI and consumer AI. I have some legislation with
several colleagues called the AI Public Awareness and Education
Campaign Act.
This seeks to provide transparency into AI, its promises,
limitations, and what consumers can expect from it. Too often,
we think about AI and our minds immediately--I think
naturally--move toward generative AI rather than the AI we have
been living with and using for the past several decades.
You are correct in your testimony that what we in Congress
do on consumer-facing AI should not impede the innovation and
deployment of technology in manufacturing like with the work
you are doing in the maritime space, for example.
As you may know, I introduced some legislation called the
Ships for America Act. This is legislation to reinvigorate
American shipbuilding, an item that, fortunately, Republicans
and Democrats alike have gotten behind.
It is something that President Trump has prioritized for
his administration, and I understand that Siemens recently
created a new maritime business to address the opportunities
the company perceived in the shipbuilding and ship repair
industries.
Can you walk me through, Ms. Ng, the benefits these types
of industrial AI capabilities can have in the effort for
America to reclaim our shipbuilding dominance?
Ms. Ng. Absolutely. Thank you for the question.
So what I would start with is by saying that if America
truly does want to reclaim shipbuilding dominance globally, we
have to outpace our competitors, not just in labor but in
building out these digital capabilities, and industrial AI is
indisputably a force multiplier here.
We think about tackling industrial AI and shipbuilding
across three different areas of a ship's lifecycle. So starting
in design, through production, and then into sustainment, we
apply applications of industrial AI across each area of that
lifecycle, everything from doing faster, more precise design
using AI-enabled simulation, through predictive maintenance,
making sure that the machines and the back shop are available
and ready to roll for every single task that is needed.
But if I were to summarize, the most important thing is
that we are using industrial AI to help sequence work so that
it is done more efficiently and more quickly, and what this
does is it reduces rework and it makes work better on the deck
plate in the shipyards because people are able to execute first
time--you know, the first time, and operate in a more
productive environment.
Thank you.
Senator Young. And as some of my constituents watch this
hearing they may think, wait a second, more productivity--that
means higher wages too, right? So that is--that is a byproduct
oftentimes of these types of investments, whether it is
yesterday's capital investments or today's AI investments.
So I do want to underscore that. What is relatedly Siemens
doing, Ms. Ng, to upskill the American workforce more broadly
to ensure workers can succeed in these high-tech, high-demand
AI-empowered roles?
Ms. Ng. Thank you for the question, Senator.
We are doing a lot, but if I were to summarize I would pick
two different things. One is we are working with an ecosystem
of partners to develop what we call micro credentials for
shipbuilding specifically but, more broadly, for manufacturing.
What this is doing is it is creating pragmatic, kind of
ready-to-use credentials that might be a day, a week, a month
long, and they prepare the workforce to go into modern
shipbuilding and start operating on day one.
The second piece is that I am incredibly proud that Siemens
has committed to training 200,000 electricians and
manufacturing experts by 2030.
We are not doing it alone. We are working with public-
private partners, universities, community colleges, and trade
programs to be able to do that, and I am extremely proud to
share that today.
Thank you.
Senator Young. Great. Thank you for your answers, Ms. Ng.
Mr. Chairman.
Senator Budd. Thank you, Senator Young.
Senator Cantwell, you are recognized.
STATEMENT OF HON. MARIA CANTWELL,
U.S. SENATOR FROM WASHINGTON
Senator Cantwell. Thank you, Chairman, and I thank you and
Ranking Member Baldwin for this important hearing, and the
witnesses for being here.
We are here today to talk about some of the most innovative
things that could help our economy, going forward, but I want
to mention we do have an NSF AI Education Act that Senator
Moran and I have introduced, and it talks about some of the
workforce issues you guys have been discussing and a Small
Business AI Training Act, that also with Senator Moran, so that
we are getting this across all aspects of our economy, and a
Future of AI Act that my colleagues Senator Young and
Hickenlooper and Blackburn and I. This is about voluntary
standards for infrastructure and innovation, mostly on safety
and security.
But the thing I wanted to focus on is the science side of
the equation. I represent a big national lab and--but the
laboratories across our country represent a unique opportunity
to take science research that basically might take you years to
do and now drive it down to months.
And so one of the things I am very worried about that we
cut 10,000 STEM Ph.D.s from the Federal agencies in 2025, so
that is not a good idea.
I represent also one of the most scientific regions of our
country and so we like scientists because they are in there
creating the next generation of economic competitiveness and
solutions and, certainly, when you look at the massive amount
of AI investment that China's making, you want to keep your
scientific workforce because they are going to help you.
But one of the things that I am most interested in hearing
from maybe you, Mr. Muro, or Dr. Shelton is how transformative
biology and chemistry lab work could be compressing it into a
few months.
As I mentioned, the Pacific Northwest Lab is doing that
already on deploying autonomous experimental platforms that you
are basically, you know, creating everything from discovery in
bio energy to advanced materials and fusion, but you are
getting it done in months instead of years.
And so I would assume that people think that this kind of
AI--I would call it AI-accelerated discovery--is, you know,
worth the commitment and worth getting this done and probably
one of the most important things that we could be doing.
Dr. Shelton or Mr. Muro, either one.
Mr. Muro. I would just say one general thing here is that
we can make alignments of our--or our computing, our AI
systems, our data, and our talent, but we have to have the
ongoing basic scientific activity functioning at a high level
as well because, really, all of that is essentially trading
data for the next iteration of all of that innovation.
So I think that is one of the things that a comprehensive
attempt to leverage AI for national good would want to look at
is making sure that you have all dimensions of the AI machinery
working, and we do think that actual science--scientific
activity and the talent it collects is absolutely central.
Senator Cantwell. Dr. Shelton.
Mr. Shelton. So logistics and warehouse work has already
seen the problem that in that industry we call islands of
automation, so you may have a conveyor belt and you may have a
ground vehicle that is carrying something and you have to get a
piece of material transferred from the autonomous ground
vehicle over to the conveyor belt.
Academic research has similar problems, although typically
with higher tech devices where you have existing automated
processes, say, a PCR machine that does DNA analysis and some
other piece of lab equipment that you have to connect together,
and robotics and specifically general purpose platforms are a
way to address that.
There is also--and this is not my field, although I do have
friends who work in this--one of the most exciting things I
think that AGI or narrow AGI specifically is able to offer up
is vastly outsized performance within a particular problem
area.
So if you are familiar with the AlphaFold project out of
DeepMind looking at protein folding, you can get AIs that are
superhumanly good at a very narrow task and that, feeding into
a lab structure that is highly automated, allows you to go from
the conceptual computation side of it down to the wet lab work
in a very compressed process.
Senator Cantwell. OK. So somebody at home, how do they
understand what you just said?
Mr. Shelton. Sure.
Senator Cantwell. I mean, I am just--I am just trying to
say, we spend a lot of money on our national labs. We have
already decided they are critical to our competitiveness as a
nation, and we are also proud of our universities.
But when I look at UW versus PNNL, we are talking about a
size--a huge size difference in the amount of research that is
done.
So now you are saying you are going to apply AI to that and
basically translate that science into faster application, as I
am saying, not years, but months, then I think this is a huge
initiative that we should be undertaking is to take all of that
research.
You just gave one example. You are basically taking one of
our big research arms and you are basically saying, let us make
sure we apply AI to it because it really is one of our most
competitive R&D efforts, right?
Mr. Shelton. One of the things, and we have seen this in
robotics research itself, is you have, you know, a human who
comes up with an idea and wants to see that translated into
something practical.
So for your question of how would I explain this to someone
who is not sitting here in the perspective of robotics land,
you have an idea. You have to get that translated into some
sort of physical reality and then you have to test that
physical reality.
I think where AI allows us to inject the best short-term
safety-controlled piece of this is to work with the human
researcher to get the math part of the process turned into
something physical that can then be tested, whether it is in
the medical space or elsewhere in industry, and allow us to
actually reduce something to practice as rapidly as possible.
Senator Cantwell. Yes, I--just the last point. I know I am
over my time.
It is that we are spending, like, about $10 billion on lab
efforts that are all about all these issues. So I am just
saying one area to double your investment is to basically say,
we are spending $10 billion here. So we have already decided we
think this is really, really important.
We are saying apply AI to that effort to make it more
efficient, because we have already decided that is where we are
spending our money, and you just gave two really good examples
of what that acceleration can deliver.
So thank you, Mr. Chairman.
Senator Budd. Thank you, Senator.
Senator Rosen, you are recognized.
STATEMENT OF HON. JACKY ROSEN,
U.S. SENATOR FROM NEVADA
Senator Rosen. Well, thank you, Chair Budd, Ranking Member
Baldwin. It is really an important hearing and I want to thank
all the witnesses for being here today.
And so I am going to start with you, Ms. Ng. I am going to
talk a little bit about AI standards and trust.
Earlier this year, I led a congressional delegation to CES
in Las Vegas with a few senators. I have been leading one
almost every year--senators from this committee. We were able
to stop by the Siemens booth, hear how Siemens is innovating on
AI, other new tools, and I am excited you are here to testify
today.
In your testimony, you outlined how using AI in advanced
manufacturing has a potential for enormous benefits. However,
the financial risk of an inaccurate outcome from an unreliable
AI tool is still very high.
Therefore, your customers have to have a clear incentive to
ensure AI tools that they integrate with their products are
trustworthy and reliable.
And so what standards of trust and reliability are your
customers demanding? Are there best practices or standards that
this Center for AI Standards and Innovation should consider
from the industry?
Ms. Ng. Absolutely, Senator, and I am so glad that you got
to visit CES. It is always a wonderful event.
To answer your question, the most important thing in
defining trust is having really strong underlying data that
informs the industrial AI application.
So what that means is being able to tap into machine-
generated operational data on a shop floor, manufacturing
systems data, and digital engineering datasets that all come
together to be able to support analysis and recommendations and
generate insights.
So what is interesting is that that is a huge piece of my
business today at Siemens is supporting our customers and
delivering what we call the sort of digital authoritative
backbone for engineering.
That serves as the foundation of which all operational
capabilities are built on top. Thank you.
Senator Rosen. Thank you. And, you know, I really do
believe, and I am sure you do too, without clear Federal
standards and guardrails, AI tools that are sold to consumers,
small businesses, large businesses, but small businesses in
particular, they have less, if any, leverage. They may not be
safe and reliable.
They could have potential to cause significant harm. So we
do have to be sure that we are paying attention. And so I want
to talk about small business and AI adoption because many
companies have prioritized AI adoption, trying to find where
this tool can improve efficiency, add value to their business,
right? How do you grow?
However, we need to ensure that the potential benefits, as
we are seeing from AI, do not continue to disproportionately
benefit only the largest corporations that have the capital and
the influence, really, to adopt safe AI and large language
models and all the other things that go along with it.
So I want to stay with you, Ms. Ng. You have a range of
customer sizes, large legacy brands like PepsiCo to very small
businesses, startups.
What is the biggest challenge you would say that your
smaller customers face adopting AI? And what can Congress do as
we think about--I am also on the Small Business and
Entrepreneurship Committee so I want to think about the large
businesses and the small.
What can Congress do to ensure that the AI adoption gap it
just does not widen, leaving our smaller innovators potentially
behind?
Ms. Ng. Absolutely, Senator, and this is an incredibly
important question.
We think about adoption for small and mid-sized businesses
in three ways. It is awareness, ability, and willingness.
So awareness is being able to communicate and understand
the art of the possible, of truly what are the business
outcomes that industrial AI can unlock for that business.
Ability is, yes, to your point, sometimes capital, but it
is also the aligned incentives with both leadership and, as
Ranking Member Baldwin mentioned, the ability to also have the
factory floor workers involved in the process.
The third piece is willingness, and this is my biggest
passion in life is organizational change management. So making
sure that there is an intentional rollout plan to all segments
of the workforce to ensure that that adoption happens.
Senator Rosen. You set me up for--I only have a few seconds
for my last question because your passion is going to lead me
to Mr. Muro to ask this question.
AI literacy, right? Because you are enthusiastic about
getting it out to everyone, but can we talk about how AI
literacy closes that AI adoption gap that we want to happen?
So thank you for setting me up for that question.
Mr. Muro. I would just say that this is where regional
ecosystems can be extremely helpful. Regional learning
ecosystems, regional technology ecosystems, and as part of that
work to make sure that adoption does include these features.
More and more regions are beginning to find their own level
on these technology solutions, and I think the Federal
Government has historically, you know, had a role in supporting
regional economic development.
I think we can be more pointed about that and maybe making
this a central area, because the talent piece--and the talent
piece are intimately connected to the--to the kind of broad
adoption that we are talking about technologically.
Senator Rosen. Well, thank you. Thank you again for all of
your work for being here.
Great hearing. Thank you, Mr. Chair.
Senator Budd. Thank you, Senator Rosen.
Ranking Member Baldwin, you are recognized for additional
questions.
Senator Baldwin. Thank you. Senator Rosen has cued me up
for my next question and my last question as we do a quick
second round before closing.
Mr. Muro, I appreciate your highlighting worker security in
your testimony. In Wisconsin, we are very proud of being the
first in the country to pass a law relating to apprenticeships
and registered apprenticeships. We did that in 1911.
The U.S. did not follow suit until 1937, but with the first
apprenticeship law in the country we continue to break
enrollment records in part because our apprenticeships focus on
in-demand fields.
I would like it if you could talk a little bit more about
the opportunities to integrate artificial intelligence into
registered apprenticeships and how these programs can help with
retraining workers in an AI world?
Mr. Muro. I mean, first, I think our view of
apprenticeships is highly shaped by its industrial past, and
that is an incredibly important theme and a place that it can
be utilized and merged with AI.
AI can also add to the learning and training dimensions of
apprenticeship through tutorials, AI--AI online supports, and
all of those.
But I think that we should think of apprenticeship as
especially a way to get hands-on work experience, and we are
going to need that more and more because it is going to become
clearer that strictly, you know, higher education degrees are
not going to be maybe more--maybe more vulnerable to change
than very specific subject matter, hands-on experience with
people learning a technology.
So I think that this is an extremely important dimension
that you have put your finger on here because in some ways,
apprenticeship points exactly at a--at a gap that I think we
are going to face in especially launching careers and setting
up pathways for development.
Senator Baldwin. Thank you.
Senator Budd. Thank you, Ranking Member Baldwin.
Mr. Giannikopoulos, there is--since you are in the
radiology world, there was a famous prediction made by the
father of AI. I am sure you are familiar with it.
I think it is Geoffrey Hinton. He said that with the rise
of AI there would be no more need for radiologists, but someone
as smart as he was is completely wrong, and your company, your
work, just proves that wrong every day.
How could someone as smart on AI as he was originally get
that sort of prediction wrong? And how could we use that as
sort of wisdom for the future as we think about the application
of AI in--personally and commercially?
Mr. Giannikopoulos. As we look at all of this, all of this
transformation, it is not unique to radiology. That was a field
that was early identified as potential for, you know,
automation and replacement. That did not work out.
It is part of the way the AI is trained. You know, it is
trained on this medical information that is generated by a
radiologist through an interpretation process. That judgment
cannot be replaced. There are going to be edge cases. There are
going to be zebras.
Again, my personal diagnostic journey, I do not fit any
algorithmic assessment that would normally get to that
diagnosis, which is why we need the human directly involved.
Like, medicine is both a science and an art, being able to put
that together and synthesize it.
Now, what AI offers as an opportunity is the ability for
the radiologists, the clinicians, to synthesize more,
understand more, access it without having to dig through
records and, you know, all these complicated systems, but
instead see it presented to them in a really easy way so that
they can make that determination.
And if you look at rural health care in particular, that is
an area where we need to augment those institutions, and in
North Carolina, ARA Health Specialists in Asheville where they
were originally based out of, they cover most of western North
Carolina right now and serve as the safety net to make sure
that patients do not have to be flown over mountains literally,
you know, to get to other health care systems.
They do that by early adoption of technology. I met them 20
years ago on the documentation side, and in my last company
they were the first--some of the first in the states to adopt
image analysis for improvement of understanding.
Now they are using new documentation tools like ours to be
able to augment that and integrating the Radiology Society of
North America evidence directly into that to really speed it
up.
So I cannot comment on how Jeff got that quite so wrong,
but I can share he did get it wrong and today, they also said
self-driving cars would take over by this time, too.
Radiologists are still driving to work in their cars.
Senator Budd. We are still waiting on that one.
I just want to thank each of you and thank the--those that
came and asked great questions. I thank the Ranking Member and,
again, each of your companies and allowing time for you all to
be here today.
Senators will have until the close of business on March 10
to submit questions for the record. The witnesses will have
until the close of business on March 24 to respond to those
questions.
This concludes today's hearing. The Committee stands
adjourned. Thank you.
[Whereupon, at 11:56 a.m., the hearing was adjourned.]
A P P E N D I X
Prepared Statement of Derek Monson, Executive Director,
Sutherland Institute
25 Ways AI is Helping People Flourish
I. Introduction: A Pro-Human Framework for Artificial Intelligence
Chairman Budd, Ranking Member Baldwin, and members of the
Subcommittee, thank you for the opportunity to submit this statement
regarding the intersection of Artificial Intelligence (AI) and human
flourishing. While much of the public debate surrounding AI focuses on
controversy and risks, AI is already creating concrete improvements to
human lives behind the scenes.
Drawing from the Sutherland Institute's recent report, 25 Ways AI
is Helping People Flourish (February 2026), authored by Ford Copple and
myself, my statement emphasizes a pro-human approach to AI policy. As
articulated by Utah Governor Spencer Cox during the recent Utah AI
Summit, such an approach empowers workers with better tools,
strengthens communities through accessible innovation, and enables
problem-solving at an unprecedented scale. Below, I outline 25 specific
examples where AI is currently improving life in education, government,
health care, and family life. It is vital that AI regulation does not
jeopardize these proven benefits, as it is contradictory to call
safeguards pro-human if they eliminate the very tools helping Americans
today.
II. Education: Empowering Teachers and Students
AI innovations are helping teachers prioritize their time more
efficiently and effectively while aiding students in far-reaching ways.
1. Reducing Burdens on Teachers: Tools like SchoolAI in Utah help
teachers spend less time on administrative duties and more time
focusing on instruction. This shift allows educators to give
more personalized attention to their students' unique needs.
2. Protecting Student Safety: New monitoring systems powered by AI
automatically detect firearms on campuses through security
camera feeds. By enhancing situational awareness, these tools
provide an additional layer of protection against school
shootings and enable faster emergency response.
3. Personalizing Student Learning: AI tools are adapting
instruction to each student's unique strengths, weaknesses, and
pace. This technology creates personalized learning plans at
scale, which increases student engagement and improves overall
outcomes.
4. Helping with Learning Disabilities: Software such as Dysolve AI
is specifically designed to help students with dyslexia improve
their reading skills. Local parents have noted that this kind
of support levels the playing field for children with learning
disabilities.
5. Increasing Student Engagement: Tools like MagicSchool AI allow
students to personally engage with historical figures, bringing
history to life. These unique lessons allow for a level of
personal interaction with curriculum that was previously
impossible.
6. Improving Reading and Language Skills: AI tools like Amira are
improving the fluency of bilingual students. Students who do
not speak English as their first language can interact with
these tools to practice and refine their reading and speaking
skills.
III. Government: Efficiency and Integrity
AI offers innovations that create efficiencies that save taxpayer
dollars and make essential government services more effective.
7. Saving Taxpayer Dollars on Infrastructure: AI-generated
``digital twins'' allow public infrastructure projects to be
tested under simulated scenarios before construction begins.
This predictive capability reduces costs and improves
efficiency by an estimated 20 percent to 30 percent.
8. Reducing Wasted Time: AI tools reduce routine administrative
tasks, allowing government employees to save up to two weeks of
time annually. This frees civil servants to focus on high-value
work rather than paperwork.
9. Improving Election Integrity: AI helps election officials flag
errors in processes and simulate various security scenarios.
These applications strengthen election security while
simultaneously reducing unnecessary waste.
10. Better Emergency Response: In Utah, AI helps screen non-
emergency calls to speed up 911 response times. This ensures
that emergency operators can stay focused on life-threatening
situations while other needs are addressed.
11. Faster Approvals while Protecting Public Safety: The FDA uses
the AI tool Elsa to streamline clinical protocol reviews and
reduce the duration of scientific assessments. This allows
lifesaving and life-improving treatments to reach the market
quicker without sacrificing patient safety.
12. Cutting Government Red Tape: Cities like Portland use AI to ease
the burden of complex permitting requirements for citizens and
businesses. This saves significant time for both the public and
government employees by simplifying bureaucratic hurdles.
IV. Health Care: Better Lives and Better Outcomes
AI is helping improve and save American lives, putting citizens in
the driver's seat for their own health, providing better health
information, and helping practitioners make more informed decisions
about care.
13. Treating Chronic Diseases: A new Utah pilot program uses AI to
speed up prescription renewals for chronic disease patients.
Simultaneously, the system detects dangerous medication
interactions to ensure patient safety.
14. Better Lives for Prosthetic Patients: Researchers at the
University of Utah have integrated AI with prosthetic limbs to
create ``bionic'' limbs. These advancements help patients
perform everyday tasks, such as reaching and gripping objects,
in more natural ways.
15. Saving Lives from Breast Cancer: AI detects breast cancer in
mammograms with increased speed and accuracy. Because early
detection is a critical variable in survival rates, this
technology is a literal lifesaver.
16. Quicker Treatment for Debilitating Disease: The Cleveland Clinic
Genome Center uses AI to detect Parkinson's Disease using
genetic, proteomic, and pharmaceutical datasets. This allows
doctors to spot early warning signs far earlier than was
previously possible.
17. Upgrading Heart Disease Diagnosis: University of Utah Health and
Intermountain Health successfully used AI to predict heart
disease onset and outcomes. This is especially helpful for
analyzing large health data records that are traditionally
difficult for humans to process.
18. Faster Skin Cancer Treatment: Doctors use AI algorithms that can
scan for skin cancer in five minutes with 99.9 percent
accuracy. This dramatically increases the speed of diagnosis
and leads to much faster treatment for patients.
19. Improving Recovery from Injuries: Apps like Wound Assistant
allow patients to monitor their own healing process from home.
Additionally, AI wearable devices using miniature cameras are
being developed to rapidly improve the healing process for
physical injuries.
V. Family Life: Support and Connection
While family policy is often fiercely debated, we must not lose
sight of the benefits AI is already offering to parents and families.
20. Connecting Loved Ones with Language Barriers: AI-generated sign
language avatars and interpreters help the deaf and hearing-
impaired connect with family members. Generative AI translators
are assisting over 70 million sign language users globally in
communicating with loved ones.
21. More Time to Spend with Children: Virtual assistants like
Ohai.ai help busy parents manage daily tasks like school
schedules. This technology frees up parents to spend more
quality time with their children.
22. Better Customer Service: Companies like Remi use AI to simplify
the process of replacing a home's roof. Customers can scan
their roof with software and receive a quote for repairs almost
immediately.
23. Saving Money, Boosting Home Values: Tools like Neighborbrite
generate custom landscaping designs tailored to a homeowner's
specific environment and taste. This reduces both the cost and
the hassle of home improvement projects.
24. Healthier, Quicker Family Meals: AI tools craft meals catered to
a family's specific dietary requests and restrictions. These
tools simplify the often-complex daily question of ``what's for
dinner'' for busy working families.
25. Improving Child Safety: Smart baby monitors like Monai use AI to
ensure infants are safe and healthy in their cribs. The monitor
sends alerts to a parent's phone if it detects concerns like a
covered face.
VI. Conclusion: A Call for Balanced Innovation
As a society, we are in the initial stages of a civilization-
altering technological change. While controversial uses draw attention,
AI is quietly and consistently improving our lives behind the scenes,
and promising even bigger life improvements in the future. The
regulatory decisions made today will determine whether AI remains a
source of human flourishing or becomes a generational lost opportunity.
We urge the Subcommittee to follow a pro-human path that protects
against harm without stifling the many essential ways AI is helping
Americans thrive, both today and tomorrow. Thank you for the
opportunity to submit this written statement.
______
Associated Builders and Contractors
Washington, DC, March 3, 2026
Hon. Ted Budd,
Chair,
Senate Committee on Commerce, Science, and Transportation,
Subcommittee on Science, Manufacturing, and Competitiveness,
U.S. Senate,
Washington, DC.
Hon. Tammy Baldwin,
Ranking Member,
Senate Committee on Commerce, Science, and Transportation,
Subcommittee on Science, Manufacturing, and Competitiveness,
U.S. Senate,
Washington, DC.
Dear Chairman Budd, Ranking Member Baldwin and Members of the U.S.
Senate Commerce, Science, and Transportation Subcommittee
on Science, Manufacturing, and Competitiveness:
On behalf of Associated Builders and Contractors, a national
construction industry trade association representing 67 chapters, more
than 23,000 member companies and millions of construction workers, I
thank you for holding this important hearing to examine how artificial
intelligence can enhance safety, increase productivity and improve care
across industries. As the voice of America's merit shop construction
industry, ABC and our contractor members are committed to building the
Nation's infrastructure safely, efficiently and responsibly. AI and
emerging construction technologies are transforming the way we work,
empowering contractors to deliver projects faster, safer and more
efficiently than ever before.
ABC launched its Technology and Innovation Initiative in 2020 to
integrate construction technology into every facet of our strategic
priorities, particularly industry-leading safety, total human health
and workforce development. Through this initiative, ABC is helping
contractors thoughtfully assess, adopt and leverage technology
solutions that strengthen safety performance, address labor shortages
and manage compressed schedules.
At ABC, safety is the foundation of everything we do. AI and
advanced technologies are reshaping jobsite safety management in
profound ways. Contractors are digitizing safety inspections, audits,
checklists and incident reports, replacing paper-based systems with
real-time, data-driven tools. By tracking both leading and lagging
safety indicators, contractors can identify risks earlier and intervene
before incidents occur. According to ABC's 2025 Health and Safety
Performance Report, companies that track these indicators experience a
62 percent reduction in total recordable incident rates and a 65
percent reduction in Days Away, Restricted, or Transferred rates
compared to those that do not. These measurable improvements show that,
when technology is integrated into a strong safety culture, it leads to
safer jobsites and ensures more workers return home safely each day.
Beyond safety, AI and related technologies are transforming
productivity and cost control. Labor and productivity tracking systems
provide real-time visibility into workforce deployment and materials
usage, enabling contractors to manage resources more effectively and
protect thin margins. Jobsite monitoring tools, including 360-degree
imaging and drone-based documentation, allow owners, general
contractors and subcontractors to collaborate with unprecedented
transparency.
Paul Hedgepath, director of virtual construction for ABC member MJ
Harris, serves as chair of ABC's Construction Technology and Innovation
Committee. He explained that one of the industry's biggest challenges
is simply accessing the information contractors already have. ``The
data is there--but it is fragmented and time-consuming to retrieve,''
he noted, pointing to the thousands of requests for information,
submittals, specifications, schedules and safety documents stored
across multiple systems on a typical project. Secure, construction-
specific AI platforms allow field leaders to ask direct questions
within their own project data and receive source-linked answers in
seconds. ``This is not automation replacing people. It is decision
support that reduces search time and increases clarity,'' Hedgepath
added. With faster access to verified information, superintendents can
prevent mistakes before work begins, and project managers can shift
time away from document hunting toward risk management and execution,
improving both safety and productivity when deployed with proper
governance and clear use cases.
Through ABC's Tech Alliance--a curated group of leading
construction technology companies--and our Tech Marketplace, we provide
contractors, most of which are small businesses, access to cost-
effective digital solutions. These partnerships deliver tools that
support bidding, project management, safety analytics, workforce
management and field collaboration. Our annual Tech Reports, including
insight papers from Hensel Phelps and Dodge Construction Network,
provide case studies and forward-looking insights to help contractors
evaluate and implement AI responsibly.
With innovation comes responsibility. AI evolves rapidly, and
without clear guidance, its use can introduce risks, including data
privacy concerns, misinformation and bias. ABC encourages contractors
to establish clear AI usage policies that ensure technologies are
deployed safely, ethically and in compliance with applicable laws.
Congress plays an important role in fostering an environment where
innovation can thrive while protecting workers, taxpayers and national
security interests. Policymakers can support responsible AI adoption in
construction by promoting regulatory clarity and ensuring that small
businesses have access to the tools and training necessary to compete
in a digital economy.
AI is not replacing the skilled men and women of the construction
industry; it is empowering them. By augmenting human expertise with
real-time data, predictive insights and intelligent automation, AI
helps contractors complete projects on time, reduce costs and, most
importantly, protect the health and safety of their workforce. The
result is stronger infrastructure, greater productivity and safer
jobsites across America.
ABC and our members stand ready to work with Congress to advance
policies that support innovation and strengthen the construction
industry's ability to build safely, ethically and efficiently.
Sincerely,
Kristen Swearingen,
Vice President, Government Affairs.
______
UVEye
Teaneck, NJ, March 3, 2026
Hon. Tedd Budd,
Chairman,
Subcommittee on Science, Manufacturing, and Competitiveness,
Committee on Commerce, Science, and Transportation,
United States Senate,
Washington, DC.
Hon. Tammy Baldwin,
Ranking Member,
Subcommittee on Science, Manufacturing, and Competitiveness,
Committee on Commerce, Science, and Transportation,
United States Senate,
Washington, DC.
Dear Chairman Budd and Ranking Member Baldwin:
On behalf of UVeye, a New Jersey-based developer and manufacturer
of advanced vehicle inspection systems, I am writing to commend the
U.S. Senate Committee on Commerce, Science, & Transportation
Subcommittee on Science, Manufacturing, and Competitiveness for its
upcoming hearing entitled, ``Less Hype, More Help: AI That Improves
Safety, Productivity, and Care.'' As a company focused on vehicle
safety, we believe AI-driven technologies can bring tangible benefits
in safety and productivity to the American people and help preserve and
improve the safety of vehicles nationwide.
UVeye is dedicated to creating safer vehicles and safer roadways
for drivers, passengers, and pedestrians, by providing objective,
consistent, and instant evaluations of the condition of a vehicle that
help identify issues before they become safety hazards. Our patented
AI-driven system uses 360+ imaging to scan each vehicle in seconds,
detecting under-body damage, tire wear, exterior dents or scratches,
alignment issues, and windshield damage. The system operates in all
weather conditions--rain, snow, or mud--and delivers standardized,
easy-to-read inspection reports. UVeye systems are installed in
Original Equipment Manufacturing (OEM) facilities, auctions,
dealerships, and heavy-duty/commercial fleets, scanning more than 2.5
million vehicles each month. This includes customers testing and
deploying autonomous vehicles (``AVs''), another AI-enabled technology.
Ensuring Vehicle Safety
UVeye's technology is helping make vehicles safer and more
efficient by identifying issues that may be invisible during manual
inspections--such as leaks, under-body damage, and worn or outdated
tires--that can lead to breakdowns or accidents. UVeye provides an
``always on'' inspection process to substantially increase the
likelihood of detecting safety issues. Faster inspections also keep
vehicles on the road and mission ready.
Protecting Consumers
For new-and used-vehicle buyers, UVeye provides transparent,
documented condition reports reducing ``hidden defect'' risk and
liability disputes. At vehicle auctions, the system reduces disputes
over damage by providing visual proof and standardized inspection data.
By improving the accuracy and consistency of vehicle inspections, UVeye
supports fairer pricing, improved confidence in used-vehicle markets,
and fewer ``surprise'' repair costs for consumers.
Boosting Efficiency
UVeye's automation reduces the burden of manual inspection,
resulting in faster, more cost-efficient inspections. Earlier detection
of damage or malfunctions means fewer breakdowns, less downtime, and
lower warranty and repair costs, generating savings for businesses and,
ultimately, consumers. When scaled across commercial fleets, AV
operators, and OEM supply chains, UVeye's technology yields multiplier
benefits in safety, transparency, and cost-efficiency.
These are just some of the ways UVeye's AI-enabled technologies are
benefiting Americans across the economy. As the Subcommittee considers
how it can support the further development and deployment of AI-enabled
systems, we urge you to consider more real-world success stories like
ours, that demonstrate the true potential of AI technologies.
Thank you again for holding this critical hearing. Please do not
hesitate to contact our counsel, Ariel Wolf ([email protected]) if you
have any questions on this letter, or if there is any further
information we can provide to the Subcommittee. We look forward to
engaging with members of the Subcommittee on this and other issues as
it continues its work to support innovative and tangible uses of AI
technologies.
Sincerely,
UVeye.
______
March 3, 2026
Hon. Ted Budd, Chairman,
Subcommittee on Science, Manufacturing, and Competitiveness,
Committee on Commerce, Science, and Transportation,
United States Senate,
Washington, District of Columbia.
Hon. Tammy Baldwin, Ranking Member,
Subcommittee on Science, Manufacturing, and Competitiveness,
Committee on Commerce, Science, and Transportation,
United States Senate,
Washington, District of Columbia.
RE: Subcommittee hearing, ``Less Hype, More Help: AI That Improves
Safety, Productivity, and Care''
Dear Chairman Budd and Ranking Member Baldwin,
ACT | The App Association (ACT) appreciates the opportunity to
submit this Statement for the Record for the Senate Committee on
Commerce, Science, and Transportation's Subcommittee on Science,
Manufacturing, and Competitiveness hearing titled ``Less Hype, More
Help: AI That Improves Safety, Productivity, and Care.'' This hearing's
focus on concrete uses of AI that improve safety, productivity, and
care is timely. Americans most often encounter AI incrementally through
practical improvements to existing tools, such as streamlining
workflows, enabling accessibility, improving diagnostics, optimizing
logistics, and strengthening cybersecurity.
Small businesses are leading the way on AI. As some of the leading
consumers, developers, and adapters of AI tools, ACT members have a
major stake in how policymakers view AI markets. ACT represents an
ecosystem valued at approximately $1.8 trillion domestically,
supporting 6.1 million American jobs.\1\ ACT members are innovators
that create the software bringing your smart devices to life. They also
make connected devices that are revolutionizing healthcare,
agriculture, public safety, financial services, and virtually all other
industries. We are concerned that state-level efforts to regulate AI
technologies before the risks and the benefits of their use are fully
understood could unnecessarily preempt ACT members' ability to compete
in AI markets and leverage the technologies.
---------------------------------------------------------------------------
\1\ https://actonline.org/wp-content/uploads/APP-Economy-Report-
FINAL-1.pdf.
---------------------------------------------------------------------------
AI is an evolving constellation of technologies that enables
computers to simulate elements of human thinking, including learning
and reasoning. In practice, Americans most often encounter AI
incrementally through improvements to existing digital services, such
as streamlined workflows, image analysis, voice recognition, and
predictive analytics. We urge policymakers to recognize these
applications as ``narrow'' AI, meaning systems that already deliver
significant societal benefits and are widely deployed by small and
medium-sized innovators. Examples of these ``narrow'' AI applications
include detecting financial fraud, strengthening cybersecurity,
enabling accessibility tools for people with disabilities, improving
healthcare diagnostics, and supporting more efficient infrastructure
and resource management. These ``narrow'' and applied AI systems
already deliver significant benefits in the real world. The policy
challenge is to ensure that governance approaches remain risk-based,
interoperable, and grounded in existing legal guardrails, while
supporting the infrastructure, standards, and workforce needed for
responsible deployment at scale.
ACT urges the Subcommittee to keep the following principles in
mind.
Existing Law Already Provides Strong Guardrails Against Harmful Conduct
A wide range of Federal and state laws already prohibit harmful
conduct regardless of whether AI is involved,\2\ and there is no AI
exemption. Section 5 of the Federal Trade Commission Act\3\ prohibits
unfair or deceptive acts or practices (UDAP), and state consumer
protection statutes apply similar standards to digital services. These
authorities reach a broad set of practices relevant to AI-enabled
products and services, including misleading claims about safety or
efficacy, failing to mitigate reasonably foreseeable risks, or
designing systems in ways that create foreseeable harms. Together,
these existing frameworks already provide meaningful guardrails for AI
developers and deployers.
---------------------------------------------------------------------------
\2\ How Existing Laws Apply to AI Chatbots for Kids and Teens,
Georgetown Law Institute for Technology Law & Policy (Nov. 10, 2025),
https://www.law.georgetown.edu/tech-institute/insights/how-existing-
laws-apply-to-ai-chatbots-for-kids-and-teens/.
\3\ Federal Trade Commission Act Sec. 5(a), 15 U.S.C. Sec. 45(a)
(prohibiting ``unfair or deceptive acts or practices in or affecting
commerce'').
---------------------------------------------------------------------------
Congress could take meaningful action by passing a comprehensive
Federal data privacy law, laying out a single set of rules for how
companies should handle consumer data. In the absence of a Federal
privacy law, we urge Congress and Federal agencies to continue
evaluating the application of these longstanding authorities before
considering new legal structures. To support a more grounded approach,
ACT launched a research initiative examining how existing federal,
state, and local laws already apply across common AI use cases.\4\ This
project aims to map notable current legal obligations across areas such
as civil rights, consumer protection, privacy, safety, labor,
competition, and intellectual property, demonstrating that AI systems
are already subject to extensive legal accountability. By establishing
a clear baseline of existing law, this effort aims to inform Federal
coordination, reduce regulatory duplication, and help policymakers
distinguish between genuine gaps and areas where improved guidance or
standards may be more effective than new legislation.
---------------------------------------------------------------------------
\4\ ACT | The App Association. Mapping Existing Laws to AI, https:/
/actonline.org/wpcontent/uploads/
ACT_Mapping_Existing_Laws_to_AI_Outline.pdf.
---------------------------------------------------------------------------
Use Risk-Based Governance and Voluntary Standards, Not Fragmented,
Precautionary Regulation
ACT strongly supports risk-based approaches to AI governance
aligned with recognized standards of safety, efficacy, and reliability.
Frameworks such as the National Institute of Standards and Technology
(NIST) AI Risk Management Framework (RMF) provide a flexible and
interoperable foundation for managing risks across diverse sectors and
use cases.\5\ In addition, industry-led, voluntary, consensus-based
standards are a cornerstone of U.S. technological leadership. The U.S.
approach reflected in OMB Circular A-119 has historically supported
innovation, interoperability, and global adoption while allowing
standards to evolve with rapidly changing technologies.\6\ Congress
should continue to support agency participation in voluntary standards
development and remove barriers that limit small business engagement in
these processes. A standards-forward, risk-based approach is especially
important for the types of applied AI at issue in this hearing,
including industrial systems, robotics, and health-care workflow tools.
These domains often depend on multi-layer supply chains and
interoperable technology stacks where clarity, testability, and
practical risk controls matter more than formalistic compliance.
---------------------------------------------------------------------------
\5\ National Institute of Standards and Technology. AI Risk
Management Framework. U.S. Department of Commerce, https://
www.nist.gov/itl/ai-risk-management-framework.
\6\ https://www.whitehouse.gov/wp-content/uploads/2017/11/Circular-
119-1.pdf.
---------------------------------------------------------------------------
Clarify Roles and Shared Responsibility Across the AI Value Chain
Effective governance depends on clearly understanding the roles and
responsibilities of different actors across the AI value chain, from
developers to deployers to downstream integrators and end users.
Assigning obligations based on demonstrated harms and each actor's
ability to mitigate risk promotes accountability while avoiding
misplaced burdens. To support this approach, ACT developed an AI Roles
& Interdependencies Framework aligned with the NIST AI RMF, describing
key stakeholders across development, distribution, deployment, and use,
and identifying practical responsibilities that support safety and
reliability.\7\ This shared-responsibility model reflects the reality
that risk mitigation is most effective when implemented by the actors
best positioned to identify and reduce risks at each stage of the
lifecycle.
---------------------------------------------------------------------------
\7\ ACT | The App Association. AI Roles & Interdependencies
Framework (May 2024), https://actonline.org/wpcontent/uploads/ACT-AI-
Roles-Interdependencies-Framework-final-text-May-2024-UK-English.pdf.
---------------------------------------------------------------------------
The Need for a National AI Framework to Avoid a Patchwork of State Laws
Congress should avoid a fragmented AI regulatory environment.
States are already advancing a growing number of AI-specific laws, many
of which would impose overlapping or inconsistent requirements for
disclosures, data practices, and risk assessments.\8\ This patchwork is
creating significant uncertainty for developers and deployers,
especially small businesses that lack the resources of their larger
competitors to navigate 50 different compliance regimes. Early evidence
shows that even small deviations in state requirements can trigger
substantial compliance costs, diverting limited resources from product
development to legal review, recordkeeping, and bespoke technical
implementations. Recent analyses show that state AI laws modeled on
European-style precautionary approaches can impose unknown and
potentially significant costs to companies,\9\ while delays in state-
level implementation underscore the difficulty of managing complex AI
obligations at the state level.
---------------------------------------------------------------------------
\8\ Stevens, Morgan. ``State of Confusion: How a Patchwork of AI
Laws Hurts Small Businesses and U.S. Competitiveness.'' ACT | The App
Association, Oct. 2025, https://actonline.org/2025/10/10/state-of-
confusion-how-a-patchwork-of-ai-laws-hurts-small-businesses-and-u-s-
competitiveness/.
\9\ ``The Hidden Cost of AI Regulations: A Survey of EU, UK, and
U.S. Companies.'' ACT | The App Association, Oct. 2025, https://
actonline.org/the-hidden-cost-of-ai-regulations-a-survey-of-eu-uk-and-
u-s-companies/.
---------------------------------------------------------------------------
A single national baseline is preferable to a system where the most
restrictive state rules dictate outcomes nationwide. Federal preemption
of conflicting state AI requirements, paired with state attorney
general enforcement of that Federal baseline, would give consumers
clarity, reduce compliance burdens, and ensure that small business
innovators can safely build and deploy the beneficial AI tools that
enhance safety, productivity, and care for all Americans.
Conclusion
ACT appreciates the Subcommittee's leadership in focusing on
practical AI tools that improve safety, productivity, and care.
Congress can reinforce the benefits of applied AI by emphasizing: (1)
risk-based governance grounded in existing law; (2) voluntary standards
and interoperable frameworks; (3) clarity on shared responsibility
across the AI value chain; and (4) a national approach that avoids
fragmented, inconsistent state-by-state mandates. We stand ready to
work with the Subcommittee, Federal agencies, and standards bodies to
support responsible AI deployment that delivers measurable benefits for
workers, consumers, and patients, while ensuring that small business
innovators can continue to compete and contribute to American
technological leadership.
Thank you for your consideration.
Respectfully submitted,
Graham Dufault,
General Counsel,
ACT | The App Association.
Kedharnath Sankararaman,
Policy Associate,
ACT | The App Association.
______
Prepared Statement of the National Council on Disability
Dear Chairman Budd, Ranking Member Baldwin and Members of the
Subcommittee:
I am writing as Acting Chairman of the National Council on
Disability (NCD), an independent, bipartisan Federal agency that
advises Congress, the President, and other Federal agencies on matters
affecting the lives of people with disabilities, to provide this
statement for inclusion in the written record of this Subcommittee's
hearing, ``Less Hype, More Help: AI That Improves Safety, Productivity,
and Care.'' NCD is providing this statement to make policymakers aware
of the susceptibility of artificial intelligence (AI) to develop
explicit and implicit biases about people with disabilities and to
advise policymakers on effective ways to ensure that these technologies
are developed with data sets that include people with disabilities.
While NCD's research recognizes the potential benefits of utilizing
AI to improve healthcare outcomes for people with disabilities, there
exist some vulnerabilities in these technologies that could negatively
impact the diagnosis and treatment of people with disabilities and
provide policymakers with erroneous information rather than accurate
solutions.
The ultimate goal of AI is to create machines that can make the
same decisions as humans.\1\ NCD's 2024 report, titled The Implicit and
Explicit Exclusion of People with Disabilities in Clinical Trials,
analyzed the use of AI in clinical trials.\2\ One study NCD examined
described how technologies such as AI, machine learning, and natural
language processing can be incorporated into several aspects of
clinical trial research.\3\ Some examples include data mining,
prescreening for possible participants, and automating invitations to
possible participants who have been prescreened through automation.
Academic researchers and the pharmaceutical industry are using AI to
mine and utilize data from electronic sources such as health records
and devices.\4\
---------------------------------------------------------------------------
\1\ Harrer S, Shah P, Antony B, et al., ``Artificial Intelligence
for Clinical Trial Design,'' Trends in Pharmacological Science,
2019;40(8):577-591. doi:10.1016/j.tips.2019.05.005.
\2\ National Council on Disability, ``The Implicit and Explicit
Exclusion of People with Disabilities in Clinical Trials,'' (2024)
available at: National Council on Disability | Federal report
illuminates need for disability inclusion in clinical trials.
\3\ Von Itzstein MS, Hullings M, Mayo H, et al., ``Application of
Information Technology to Clinical Trial Evaluation and Enrollment: A
Review,'' JAMA Oncology, 2021;7(10):1559-1566. doi:10.1001/
jamaoncol.2021.1165.
\4\ Woo M, ``An AI Boost for Clinical Trials,'' Nature,
2019;573(7775): S100-S102. doi:10.1038/D41586-019-02871-3.
---------------------------------------------------------------------------
Our 2024 report found that a contributing factor to health care
disparity outcomes for people with disabilities was physicians'
erroneous assumptions about the values and expectations of their
patients, assumptions that mirror widespread, stigmatized societal
views about the disabled.\5\ Due to these concerns, NCD found that
disability cultural competence should be a core strategy for the
healthcare system in order to reduce healthcare disparities for people
with disabilities.\6\ Strong evidence exists that cultural training for
health care professionals improves providers' knowledge, understanding,
and skills for treating patients from culturally, linguistically, and
socioeconomically diverse backgrounds.\7\
---------------------------------------------------------------------------
\5\ Shakespeare T, Iezzoni LI, Groce NE, ``Disability and the
Training of Health Professionals,'' Lancet, 2009;374(9704):1815-1816.
\6\ Association of American Medical Colleges, Cultural Competence
Education for Medical Students, 2005. https://www.aamc.org/download/
54338/data/culturalcomped.pdf.
\7\ Govere L, Govere EM, ``How Effective Is Cultural Competence
Training of Healthcare Providers on Improving Patient Satisfaction of
Minority Groups? A Systematic Review of literature,'' Worldviews on
Evidence-Based Nursing, 2016;13(6):402-410. Accessed January 16, 2024.
https://sigmapubs.onlinelibrary.wiley.com/doi/full/10.1111/wvn.12176.
---------------------------------------------------------------------------
While NCD's 2024 research was limited to AI and clinical trials, we
reasonably believe our findings and recommendations can be generalized
to other forms of AI technology in healthcare. Our findings were based
on published studies, legislation, and clinical trial protocols as well
as subject matter expert interviews, trial participant interviews,
healthcare provider and participant surveys as well as feedback from
stakeholders at the National Institutes of Health (NIH) and Food and
the Drug Administration (FDA).
Because AI is intended to develop the same decision-making
capabilities as humans, NCD is similarly concerned that AI technologies
may inadvertently develop the same assumptions and biases about people
with disabilities. For this reason, NCD advises policymakers to review
the use of AI in healthcare in general and establish regulations as
needed to ensure that these technologies are built on data sets that
include people with disabilities, so that implicit and explicit biases
are not accidentally developed or ``learned.''
Thank you for the opportunity to provide a brief summary of NCD's
relevant research, analysis, and recommendations on ways to improve the
use of AI technology in the healthcare system for people with
disabilities. We welcome the opportunity to brief the Committee and its
staff in depth on any of these or related topics at your direction and
request.
To that end, please do not hesitate to contact our Executive
Director, Ana Torres-Davis, [email protected], and Director of
Legislative Affairs and Outreach, Anne Sommers McIntosh,
[email protected], who will be glad to address any request for follow-
up you may have or provide a more in-depth briefing on any of our
reports and advisement.
Respectfully,
Neil Romano,
Acting Chairman,
National Council on Disability.
[all]