Trump AI Export Controls Face First Amendment Challenge

Legal questions mount over AI access restrictions

The Trump administration's recent restrictions on advanced AI models are triggering constitutional questions that mirror landmark free speech battles from the 1990s, according to legal experts and security researchers familiar with emerging challenges to the policy.

The administration this month suspended access to Anthropic's frontier AI models—Fable 5 and Mythos 5—citing security vulnerabilities. While Commerce Secretary Howard Lutnick partially restored access to Mythos 5 for select U.S. critical infrastructure and cybersecurity organizations last Friday, broader restrictions remain in place.

Legion, an AI company, filed suit in federal court last week arguing the administration exceeded its authority with the export controls, though the complaint did not explicitly invoke First Amendment grounds. According to a white-hat AI researcher speaking on condition of anonymity, additional constitutional challenges may be forthcoming.

Why it matters

The legal framework governing AI access could determine whether security researchers can study and fix vulnerabilities in systems that adversaries will inevitably obtain through other channels. The outcome may also establish precedent for how constitutional speech protections apply to emerging AI technologies that increasingly mediate access to information.

Echoes of the crypto wars

The constitutional argument draws directly from 1990s litigation over encryption software. In Bernstein v. United States, the 9th U.S. Circuit Court of Appeals ruled in 1996 that source code constitutes speech protected by the First Amendment, and that government regulations preventing its publication were unconstitutional. The Electronic Frontier Foundation successfully challenged federal export controls in that case.

Phil Zimmermann, creator of Pretty Good Privacy encryption software, spent three years under federal criminal investigation after publishing his encryption program online. Prosecutors ultimately declined to indict, but the episode helped establish that computer code receives First Amendment protection. Zimmermann told POLITICO that had his case gone to trial, "free speech would have been probably our most successful strategy for winning in court."

Tyler Tone, legislative analyst for technology and free expression at the Foundation for Individual Rights and Expression, argues similar reasoning applies to AI systems. He contends that AI models embody layers of human expression—from developers' choices about training data to users' right to receive information—placing them within established First Amendment doctrine.

The security researcher's dilemma

While acknowledging the merit of First Amendment arguments, Zimmermann believes the administration's national security rationale actually supports the opposite conclusion. Security researchers need access to the same powerful models that adversaries will obtain to identify and fix vulnerabilities before attacks occur.

"The bad guys are certainly going to get their hands on these tools and they are going to attack us," Zimmermann said. "We need these tools to identify the vulnerabilities so that we can fix them."

The researcher who spoke with POLITICO on background framed the issue as government jawboning: "Export controls are an infringement on speech because when you publish research or security vulnerabilities you're expressing yourself by accessing the technology and fixing the problems."

The Commerce Department did not immediately respond to a request for comment on the constitutional questions.

These details were first reported by POLITICO's West Wing Playbook.