The Crypto Wars: How Strong Encryption Stopped Being a Munition

Every piece on this site so far assumes something that was not true for most of the decade before it: that a person can write cryptographic software, publish it on the internet, and not thereby have exported a defense article.

That assumption was fought over for years in United States rulemaking and in three federal courts, by people who had never heard of digital cash and were not arguing about it. The outcome is the precondition for everything that follows, and it is the most consistently skipped chapter in this history. What follows is the sequence, restricted to the public record, with the disputes left as disputes.

The starting position

Cryptographic software was a munition. That is not a figure of speech and not a gloss. It was designated as a defense article on the United States Munitions List, administered by the State Department under the International Traffic in Arms Regulations pursuant to the Arms Export Control Act.

The specific citation appears in the Karn litigation below, where the Office of Defense Trade Controls stated that the item before it “is designated as a defense article under category XIII(b) (1) of the United States Munitions List”. Category XIII is where the cryptography sat.

9 February 1994. The National Institute of Standards and Technology publishes FIPS 185, the Escrowed Encryption Standard. Its abstract specifies “an encryption/decryption algorithm and a Law Enforcement Access Field (LEAF) creation method”, and states that “the algorithm and the LEAF creation method are classified and are referenced, but not specified, in the standard”. The LEAF, the document says, “is used in a key escrow system that provides for decryption of telecommunications when access to the telecommunications is lawfully authorized”.

Read the standard rather than the arguments about it and the shape is plainly on the page: strong encryption, in a device, with a mechanism for authorised access, and the algorithm withheld. FIPS 185 was withdrawn on 19 October 2015.

Karn: the book and the diskette

9 March 1994. The plaintiff in what became Karn v. U.S. Department of State submits a commodity jurisdiction request for a diskette containing the source code printed in Part Five of the book Applied Cryptography. The book itself had already been determined not to be subject to the jurisdiction of the State Department under the regulations. The identical text, on a floppy disk, was designated a defense article.

That is the whole case. The same characters, in the same order, by the same author, were a publication in one physical form and a controlled export in another.

22 March 1996. Judge Charles R. Richey, United States District Court for the District of Columbia, Civil No. 95-1812 (CRR), rules for the government. He grants dismissal of the claim brought under the Administrative Procedure Act, holding that section 2778(h) of the Arms Export Control Act precluded judicial review of the designation decision, and grants summary judgment for the defendants on the First and Fifth Amendment claims, finding the regulation content neutral and rationally related to national security interests.

The transfer

15 November 1996. Executive Order 13026, “Administration of Export Controls on Encryption Products”, signed by President Clinton. It moves encryption products that were or would be designated as defense articles in Category XIII of the Munitions List over to the Commerce Control List, administered by the Commerce Department under the Export Administration Regulations.

Two provisions are worth reading rather than summarising. The order makes sections 4(c) and 6(h)(2) to (4) of the Export Administration Act of 1979, which concern foreign availability, inapplicable to export controls on such encryption products. And section 1(c) states that publicly available encryption software shall not be considered or treated as “technology” within the Act’s definition.

That second one is the load bearing sentence and it is the opposite of what a casual reader expects from the word “publicly available”. Ordinarily, published information escapes control precisely because it is published. Section 1(c) carves encryption software out of that treatment, which is to say the transfer to Commerce was a transfer of jurisdiction and not a release. Section 3 adds that the order is not intended to create any right to administrative or judicial review.

The Sixth Circuit later summarised the move in one sentence, and this is the sentence to quote if you quote one: “In 1996, the President transferred export jurisdiction over nonmilitary encryption items from the State Department to the Commerce Department’s Bureau of Export Administration.”

Junger: what the regime looked like from a classroom

12 June 1997. Peter Junger, a law professor who taught a course in computers and the law, submits three applications to the Commerce Department for commodity classifications of encryption software programs and other items. He wanted to post encryption source code he had written on his own web site.

4 July 1997. The Export Administration tells him that Export Control Classification Number 5D002 covered four of the five software programs he had submitted. It found that the first chapter of his textbook, Computers and the Law, was an allowable unlicensed export. Its position, as the Sixth Circuit recorded it, was that the printed book chapter containing encryption code could be exported, and that export of the same text in electronic form would require a license if it contained 5D002 software.

The print and electronic distinction from Karn had survived the change of jurisdiction intact.

1998. The district court, Northern District of Ohio at Akron, Judge James S. Gwin, grants summary judgment to the government. On the Sixth Circuit’s account, it “found that encryption source code is not sufficiently expressive to be protected by the First Amendment, that the Export Administration Regulations are permissible content-neutral restrictions, and that the Regulations are not subject to a facial challenge as a prior restraint on speech”.

Bernstein: won, then unwon

6 May 1999. A three judge panel of the Ninth Circuit, Judges Fletcher, Bright and T.G. Nelson, decides Daniel Bernstein’s challenge. Bernstein was a doctoral student who had written an encryption method he called Snuffle and wanted to publish its source code; the State Department had responded that Snuffle was a munition under the regulations. The panel holds “the challenged regulations constitute a prior restraint on speech that offends the First Amendment”, on the basis that the licensing scheme burdened scientific expression, vested boundless discretion in officials and lacked adequate procedural safeguards.

The panel also wrote, in the same opinion: “We emphasize the narrowness of our First Amendment holding. We do not hold that all software is expressive.”

30 September 1999. The Ninth Circuit orders the case reheard by the en banc court pursuant to Circuit Rule 35-3, and the three judge panel opinion is withdrawn.

This is where most retellings of the crypto wars go wrong, and it is worth being blunt about it. A withdrawn opinion is not law. Bernstein is cited constantly as the case that established that code is speech, and the opinion that said so was vacated by the court that issued it, four months later, without a merits ruling replacing it. The proposition may be right. That opinion is not the authority for it.

Junger on appeal, and the regulations moving underneath

January 2000. With Junger argued and awaiting decision, the Bureau of Export Administration issues an interim final rule, Revisions to Encryption Items, 65 Fed. Reg. 2492, to be codified at 15 C.F.R. Parts 734, 740, 742, 770, 772 and 774.

The regime it produced is worth stating precisely, because “encryption was deregulated in 2000” is not what the text says. Encryption software remained controlled under ECCN 5D002 for national security reasons. A license was required for the export of all encryption items to all foreign destinations except Canada. Encryption software in printed form was not subject to the Regulations, under 15 C.F.R. section 734.3(b)(2). And for encryption software, the definition of “export” included publication of the software on the internet unless steps were taken to restrict foreign access to the site.

17 December 1999, argued. 4 April 2000, decided. The Sixth Circuit, Chief Judge Boyce F. Martin Jr. writing, No. 98-4045, reverses. The holding is one sentence:

Because computer source code is an expressive means for the exchange of information and ideas about computer programming, we hold that it is protected by the First Amendment.

And on the district court’s reasoning that function overrides expression:

The fact that a medium of expression has a functional capacity should not preclude constitutional protection. Rather, the appropriate consideration of the medium’s functional capacity is in the analysis of permitted government regulation.

The disposition was not a victory on the merits of the regulations. The court reversed and remanded “for further consideration of Junger’s constitutional claims in light of the amended regulations”, noting that after the amendments the district court should examine the new rules to determine whether Junger could bring a facial challenge at all.

And the court said this, which is the reason the last section of this piece is worded the way it is:

In the present case, the record does not resolve whether the exercise of presidential power in furtherance of national security interests should overrule the interests in allowing the free exchange of encryption source code.

What was expanding, what was contracting

What expanded through this decade was the civilian demand for strong cryptography, driven by commerce arriving on a public network. What contracted, slowly and unevenly, was the position that the government could treat a published algorithm as a weapon.

Neither curve crossed cleanly and neither crossed on a single date, which is why this is a timeline rather than an anniversary. The crossing was made out of a transfer of jurisdiction that was also a carve out, a rulemaking that liberalised and retained, and two appellate opinions of which one no longer exists.

Who could tell at the time is the part that matters for this site. The mailing list could tell, and said so continuously, and was arguing for the freedom to publish cryptographic code as a civil liberties question rather than as groundwork for anything in particular. Nobody in that fight was clearing a path for digital money. The people who actually needed the path cleared, the ones who had already tried to build private electronic cash and been stopped by distribution and by the operator being reachable, were losing on entirely different grounds at the same time.

The freedom to publish the code was won by other people, for other reasons, before the code existed. That is the finding, and it is a strange one: the single most important precondition in this history was secured by a fight that had nothing to do with it, and it was secured incompletely, and the incompleteness is still there.