Rep. Don Beyer (D-Va.) is faulting the Trump administration’s abrupt move to cut off access to two of Anthropic’s most advanced AI models, calling the federal government’s approach to the technology “haphazard.”
Anthropic, one of the country’s largest AI companies, said it was ordered this month to suspend access to its Fable 5 and Mythos 5 models for any foreign national — a directive it said it could comply with only by switching the models off for every user worldwide. The Commerce Department issued the order on June 12, citing national security, various outlets reports.
Anthropic’s other models were not affected.
ARLnow asked Beyer whether he believed the move to limit public access was justified and proportionate. In a statement, he aimed his criticism at the administration’s overall handling of AI rather than the specifics of the Anthropic order.
“The Administration’s haphazard approach to managing ever more powerful AI models is deeply concerning and does little to protect the American public from the serious risks posed by these models,” Beyer said.
He called the suspension, imposed “through last-minute export controls,” a poor template for governing the technology — “not a sustainable blueprint for safe and responsible AI governance.” The answer, he argued, lies with Congress.
“The best path forward is for Congress to establish clear, comprehensive standards for model evaluations, deployment decisions, and risk mitigation so that developers, regulators, and the public can operate within a consistent and predictable framework,” Beyer said.
He made the case a day after he and two colleagues — Reps. Deborah Ross (D-N.C.) and Jeff Hurd (R-Colo.) — introduced a bipartisan bill aimed at that kind of framework. The AI Flaw Reporting and Security Enhancement Act, introduced Tuesday (June 23) as H.R. 9333, would set up a voluntary federal program for AI developers to report vulnerabilities in their models before they can be exploited.
The measure would direct the National Institute of Standards and Technology to build a reporting system for AI flaws modeled on its National Vulnerability Database, the catalog of cybersecurity weaknesses used by organizations around the world. It would also have NIST work with the private sector on ways to detect and fix AI flaws, convene stakeholders to study the problem across industries, and report back to Congress within three years.
“Congress cannot afford to fall behind in addressing these emerging risks and ensuring appropriate safeguards are in place,” Beyer said in announcing the bill. “Our legislation would establish a centralized reporting mechanism for potential security and safety vulnerabilities in AI systems, allowing for timely and coordinated action to mitigate risks.”
Arlington’s congressman has been one of Capitol Hill’s most active voices on AI. He sits on the House’s bipartisan AI Task Force, helped lead an early push on federal AI regulation and has warned about “Skynet”-style risks from the technology. He is also pursuing a master’s degree in machine learning at George Mason University.
The two models at the center of the dispute are among the most powerful Anthropic has released.
Mythos 5, available only to the government and a limited set of corporate partners, is regarded as unusually good at finding software vulnerabilities — an ability useful to cybersecurity defenders, but one critics have warned could be turned into a cyberweapon. Fable 5, released to the public days before the suspension, is built on the same technology with safeguards meant to block those high-risk uses.
The order followed a reported method for getting around, or “jailbreaking,” those safeguards, according to Anthropic. The government has not publicly detailed its concern.
In a statement disputing the basis for the directive, Anthropic said the reported flaw was a narrow one — essentially asking the model to read a block of code and fix its bugs — and that comparable capabilities are available from other public AI models, including OpenAI’s GPT-5.5. The company called the episode a “misunderstanding” and said it is working to restore access.
Fable 5 remains unavailable to the public as of publication time.
It is far from the only friction between Anthropic and the administration. The company is already suing the federal government over a Defense Department move to designate it a national security “supply chain risk,” a label historically reserved for foreign adversaries.
