BNUMBER:  B-245714
DATE:  December 13, 1991
TITLE:  National Institute of Standards and Technology--Use
of Electronic Data Interchange Technology to Create
Valid Obligations

**********************************************************************

Matter of:     National Institute of Standards and Technology--Use of
               Electronic Data Interchange Technology to Create Valid
               Obligations

File:          B-245714

Date:          December 13, 1991

DIGEST

Contracts formed using Electronic Data Interchange technologies may
constitute valid obligations of the government for purposes of 31
U.S.C.  sec.  1501, so long as the technology used provides the same degree
of assurance and certainty as traditional "paper and ink" methods of
contract formation.

DECISION

By letter dated September 13, 1991, the Director, Computer Systems
Laboratory, National Institute of Standards and Technology (NIST),
asked whether federal agencies can use Electronic Data Interchange
(EDI) technologies, such as message authentication codes and digital
signatures, to create valid contractual obligations that can be
recorded consistent with 31 U.S.C.  sec.  1501.  For the reasons stated
below, we conclude that agencies can create valid obligations using
properly secured EDI systems.

BACKGROUND

EDI is the electronic exchange of business information between
parties, usually via a computer, using an agreed upon format.  EDI is
being used to transmit shipping notices, invoices, bid requests, bid
quotes, and other messages.  Electronic contracting is the use of EDI
technologies to create contractual obligations.  EDI allows the
parties to examine the contract, usually on video monitors, but
sometimes on paper facsimiles, store it electronically (for example,
on magnetic tapes, on discs or in special memory chips), and recall it
from storage to review it via electronic means.  Using EDI
technologies, it is possible for an agency to contract in a fraction
of the time that traditional practices take.

As NIST pointed out in its request, the "paperless" nature of the
technology has raised the question of whether electronic contracts
constitute obligations which may be recorded against the government. 
NIST is in the process of developing standards for electronic
signatures to be used in various applications,[1] including the
formation of contracts, but has been advised that section 1501 imposes
a barrier to the use of electronic technologies by federal agencies in
this regard.

DISCUSSION

Section 1501 establishes the criteria for recording obligations
against the government.  The statute provides, in pertinent part, as
follows:

     "(a) An amount shall be recorded as an obligation of the United
     States Government only when supported by documentary evidence
     of--

        (1) a binding agreement between an agency and another person
        (including an agency) that is--

          (A) in writing, in a way and form, and for a purpose
          authorized by law. . . ."

31 U.S.C.  sec.  1501(a)(1)(A).

Under this provision, two requirements must be satisfied:  first, the
agreement must bind both the agency and the party with whom the agency
contracts; second, the agreement must be in writing.

Binding Agreement

The primary purpose of section 1501(a)(1) is "to require that there be
an offer and an acceptance imposing liability on both parties."  39
Comp. Gen. 829, 831 (1960) (emphasis in original).  Hence the
government may record an obligation under section 1501 only upon
evidence that both parties to the contract willfully express the
intent to be bound.  As explained below, EDI technology provides both
the agency and the contractor the means to electronically "sign" a
contract.

A signature traditionally has provided such evidence.  See generally
65 Comp. Gen. 806, 810 (1986).  Because of its uniqueness, the
handwritten signature is probably the most universally accepted
evidence of an agreement to be bound by the terms of a contract.  See
65 Comp. Gen. at 810.  Courts, however, have demonstrated a
willingness to accept other notations, not necessarily written by
hand.  See, e.g., Ohl & Co. v. Smith Iron Works, 288 U.S. 170, 176
(1932) (initials); Zacharie v. Franklin, 37 U.S. (12 Pet.) 151, 161-62
(1838) (a mark); Benedict v. Lebowitz, 346 F.2d 120 (2nd Cir. 1965)
(typed name); Tabas v. Emergency Fleet Corporation, 9 F.2d 648, 649
(E.D. Penn. 1926) (typed, printed or stamped signatures); Berryman v.
Childs, 98 Neb. 450, 153 N.W. 486, 488 (1915) (a real estate brokerage
used personalized listing contracts which had the names of its brokers
printed on the bottom of the contract in the space where a handwritten
signature usually appears).

As early as 1951, we recognized that a signature does not have to be
handwritten and that "any symbol adopted as one's signature when
affixed with his knowledge and consent is a binding and legal
signature."  B-104590, Sept. 12, 1951.  Under this theory, we approved
the use of various signature machines ranging from rubber stamps to
electronic encryption machines ranging from rubber stamps to
electronic encryption devices.  See 33 Comp. Gen. 297 (1954);
B-216035, Sept. 20, 1984.  For example, we held that a certifying
officer may adopt and use an electronic symbol generated by an
electronic encryption device to sign vouchers certifying payments. 
B-216035, supra.  The electronic symbol proposed for use by certifying
officers, we concluded, embodied all of the attributes of a valid,
acceptable signature:  it was unique to the certifying officer,
capable of verification, and under his sole control such that one
might presume from its use that the certifying officer, just as if he
had written his name in his own hand, intended to be bound.

EDI technology offers other evidence of an intent to be bound with the
same attributes as a handwritten signature.  We conclude that EDI
systems using message authentication codes which follow NIST's
Computer Data Authentication Standard (Federal Information Processing
Standard (FIPS) 113)[2] or digital signatures following NIST's Digital
Signature Standard, as currently proposed, can produce a for of
evidence that is acceptable under section 1501.

Both the message authentication code and the digital signature are
designed to ensure the authenticity of the data transmitted.  They
consist of a series of characters that are cryptographically linked to
the message being transmitted and correspond to no other message. 
There are various ways in which a message authentication code or
digital signature might be generated.  For example, either could be
generated when the sender inserts something known as a "smart card"[3]
into a system and inputs the data he wants to transmit.  Encoded on a
circuit chip located on the smart card is the sender's private key. 
The sender's private key is a sequence of numbers or characters which
identifies the sender, and is constant regardless of the transmission. 
The message authentication code and the digital signature are
functions of the sender's private key and the data just loaded into
the system.  The two differ primarily in the cryptographic methodology
used in their generation and verification.

After loading his data into the system, the sender notifies the system
that he wants to "sign" his transmission.  Systems using message
authentication codes send a copy of the data tot he chip on the smart
card; the chip then generates the message authentication code by
applying a mathematical procedure known as a cryptographic algorithm. 
Systems using digital signatures will send a condensed version of the
data to the smart card, which generates the digital signature by
applying another algorithm, as identified in NIST's proposed standard. 
The card returns the just-generated message authentication code or
digital signature to the system, which will transmit it and the data
to the recipient.

Under either approach, when an offeror or a contracting officer
notifies the system that he wants to "sign" a contract for being
transmitted, he is initiating the procedure for generating a message
authentication code or digital signature with the intention of binding
his company or agency, respectively, to the terms of the contract.[4] 
The code or digital signature evidences that intention, as would a
handwritten or other form of signature.  Both, generated using the
sender's private key, are unique to the sender; and, the sender
controls access to and use of his "smart card," where his key is
stored.

They are also verifiable.  When the recipient receives the contract,
either on his computer monitor or in paper facsimile, it will carry,
depending on which approach is used, a notation which constitutes the
message authentication code or the digital signature of the sender,
necessary information to validate the code or the signature and,
usually, the sender's name.  The recipient can confirm the
authenticity of the contract by entering the data that he just
received and asking his system to verify the code or the digital
signature.  The system will then use the information provided by the
sender and either verify or reject it.[5]  Both approaches use a key
to verify the message just received; however, the digital signature
requires application of a different key from that used to verify a
message authentication code.  The change of any data included in the
message as transmitted will result in an unpredictable change to the
message authentication code or the digital signature.  Therefore, when
they are verified, the recipient is virtually certain to detect any
alteration.

Writing

To constitute a valid obligation under section 1501(a)(1)(A), a
contract must be supported by documentary evidence "in writing."  As
NIST pointed out, some have questioned whether EDI, because of the
paperless nature of the technology, fulfills this requirement.  We
conclude that it does.

Prior to the enactment of section 1501, originally section 1311 of the
Supplemental Appropriations Act of 1955,[6] there was no "clean cut
definition of obligations."  H.R. Rep. No. 2266, 83rd Cong., 2d Sess.
50 (1954).  Some agencies had recorded questionable obligations,
including obligations based on oral contracts, in order to avoid
withdrawal and reversion of appropriated funds.  See 51 Comp. Gen.
631, 633 (1972).  Section 1501 was enacted not to restrict agencies to
paper and ink in the formation of contracts, but because, as one court
noted, "Congress was concerned that the executive might avoid spending
restrictions by asserting oral contracts."  United States v. American
Renaissance Lines, 494 F.2d 1059, 1062 (D.C. Cir. 1974) cert. denied,
419 U.S. 1020 (1974).  The purpose of section 1501 was to require that
agencies submit evidence that affords a high degree of certainty and
lessens the possibility of abuse.  See H.R. Rep. No. 2266 at 50.

While "paper and ink" offers a substantial degree of integrity, it is
not the only such evidence.  Some courts, applying commercial law (and
the Uniform Commercial Code in particular), have recognized audio tape
recordings, for example, as sufficient to create contracts.  See,
e.g., Ellis Canning Company v. Bernstein, 348 F. Supp. 1212 (D. Colo.
1972).  The court, citing a Colorado statute, stated that the tape
recording of the terms of a contract is acceptable because it is a
"reduction to tangible form."[7]  Id. at 1228.  In a subsequent case,
a federal Court of Appeals held that an audio tape recording of an
agreement between the Gainesville City Commission and a real estate
developer was sufficient to bind the Commission.  Londono v. City of
Gainesville, 768 F.2d 1223 (11th Cir. 1985).  The court held that the
tape recording constituted a "signed writing."  Id. at 1228.

In our opinion, EDI technology, which allows the contract terms to be
examined in human readable form, as on a monitor, stored on electronic
media, recalled from storage and reviewed in human readable form, has
an integrity that is greater than an audio tape recording and equal to
that of a paper and ink contract.  Just as with paper and ink, EDI
technology provides a recitation of the precise terms of the contract
and avoids the risk of error inherent in oral testimony which is based
on human memory.[8]  Indeed, courts under an implied-in-fact contract
theory, have enforced contracts on far less documentation than would
be available for electronic contracts.  See Clark v. United States, 95
U.S. 539 (1877).  See also Narva Harris Construction Corp. v. United
States, 574 F.2d 508 (Ct. Cl. 1978).

For the purpose of interpreting federal statutes, "writing" is defined
to include "printing and typewriting and reproductions of visual
symbols by photographing, multigraphing, mimeographing, manifolding,
or otherwise."  1 U.S.C.  sec.  1 (emphasis added).  Although the terms of
contracts formed using EDI are stored in a different manner than those
of paper and ink contracts, they ultimately take the form of visual
symbols.  We believe that it is sensible to interpret federal law in a
manner to accommodate technological advancements unless the law by its
own terms expressly precludes such an interpretation, or sound policy
reasons exist to do otherwise.  It is evident that EDI technology had
not been conceived nor, probably, was even anticipated at the times
section 1501 and the statutory definition of "writing" were enacted. 
Nevertheless, we conclude that, given the legislative history of
section 1501 and the expansive definition of writing, section 1501 and
1 U.S.C.  sec.  1 encompass EDI technology.

Accordingly, agencies may create valid obligations using EDI systems
which meet NIST standards for security and privacy.

Comptroller General
of the United States

1. The Congress has mandated that NIST (formerly the National Bureau
of Standards) establish minimum acceptable practices for the security
and privacy of sensitive information in federal computer systems. 
Computer Security Act of 1987, Pub. L. No. 100-235,  sec.  2, 101 Stat.
1724 (1988).

2. FIPS 113 adopts American National Standards Institute (ANSI)
standard X9.9 for message authentication.  It outlines the criteria
for the cryptographic authentication of electronically transmitted
data and for the detection of inadvertent and/or intentional
modifications of the data.  By adopting the ANSI standard, FIPS 113
encourages private sector applications of cryptographic
authentication; the same standard is being adopted by many financial
institutions for authenticating financial transactions.

3. A smart card is the size of a credit card.  It contains one or more
integrated circuit chips which function as a computer.

4. NIST officials advise us that technology using message advise us
that technology using message authentication codes and digital
signatures will be available to both contractors and contracting
officers for use in government contracting.

5. For the sake of simplicity, this example does not describe the
complicated system of controls used to ensure that (1) no human knows
the sender's private key and (2) the information received from the
sender for validating the message authentication code or digital
signature is correct and accurate.

6. Pub. L. No. 663, 68 Stat. 800, 830 (1954).

7. Other courts, interpreting the laws of other states, have held that
a tape recording is not acceptable.  See Sonders v. Roosevelt, 102
A.D.2d 701, 476 N.Y.S.2d 331 (1984); Roos v. Aloi, 127 Misc.2d 864,
487 N.Y.S.2d 637 (N.Y. Sup. Ct. 1985).

8. Of course, just as with any contract or other official document, an
agency must take appropriate steps to ensure the security of the
document, for example, to prevent fraudulent modification of the
terms.  Agencies should refer to NIST standards in this regard.  See,
e.g., FIPS 113 (regarding message authentication codes).  In addition,
agencies should refer to the GSA regulations regarding the maintenance
of electronic records, see 41 C.F.R.  sec.  201-45.2, and to the Federal
Rules of Evidence with regard to managing electronic records to ensure
admissibility, see generally Department of Justice Report,
"Admissibility of Electronically Filed Federal Records as Evidence,"
Systems Policy Staff, Justice Management Division (October 1990).