The Trump administration has forced a major artificial intelligence company to take its most advanced cybersecurity models offline, marking an unprecedented government intervention in deployed AI technology and raising concerns among security experts that the move could weaken America's defensive capabilities while damaging trust in U.S. technology abroad.
Anthropic, a San Francisco-based AI developer, complied Friday with a Commerce Department directive that barred non-Americans—including the company's own employees—from accessing its latest models, Fable 5 and Mythos 5. The company said it did not believe the government's action was warranted by the security concerns it had flagged about the technology.
The shutdown comes after tensions between the Trump administration and Anthropic over the company's efforts to build safeguards into AI development. Defense Secretary Pete Hegseth sought to designate Anthropic as a supply chain risk following a Pentagon contract dispute, an unprecedented move against a U.S. company that Anthropic has challenged in two federal courts. The company said it wanted assurance the Pentagon would not use its technology in fully autonomous weapons or surveillance of Americans. Hegseth said the company must allow for any uses the Pentagon deemed lawful.
The Security Paradox
More than 100 cybersecurity experts and leaders from companies including Adobe and Nvidia issued a letter Sunday urging the U.S. government to lift the export control directives and commit to "an open, scientific and transparent process of handling AI risk assessments in the future." The experts said Anthropic's Mythos models are "quite good" at finding software flaws and weaponizing exploits, but are "not uniquely good at these tasks," and that many organizations regularly use other foundation and open-source models for the same security work.
The letter warned that removing advanced cybersecurity capabilities from network defenders in the U.S. is "dangerous" when American adversaries are rapidly advancing their own AI systems. The experts noted that China's models are "only months behind the best American models," and that it is "even likely China's government has access to private capabilities beyond what has been made public."
Katie Moussouris, a cybersecurity veteran and researcher who founded Luta Security, stated that the bypass described by security researchers "should never have triggered an export control." She argued that "the behavior described in the paper cannot meaningfully be fixed, and any attempt would only weaken the model for defense."
Precedent for Government Overreach
The intervention represents the U.S. government's most significant step yet to restrict access to advanced AI models. It came 10 days after President Donald Trump signed an executive order establishing a framework for the federal government to vet national security risks of the most advanced AI systems for up to a month before their public release. The executive order specified that participation by AI developers would be voluntary.
However, the Commerce Department's directive to Anthropic operated without such voluntary participation—the company was ordered to comply. TechCrunch reported that the enforcement letter effectively forced the company to pull its latest AI models offline and showed that the AI industry is not immune to government interference. The outlet noted that the government's action appeared to require no court approval.
Justin Hendrix, editor of Tech Policy Press, warned that the Trump administration's move is "likely to raise alarms in foreign capitals about the reliability of American AI for critical applications." He said the message sent is that AI companies in the United States cannot be trusted to operate without interference from the U.S. government.
The Washington Post reported that pressuring Anthropic to retract its Fable model has established a potential precedent for intervening in deployed AI models through export control authorities. The Post also noted that Amazon CEO Andy Jassy first raised concerns about a Fable jailbreak with the White House, suggesting that Amazon's strategic partnership with Anthropic was becoming more complex.
Joe Khawam of the Law Reform Institute was quoted as saying the administration is likely to enforce export controls on AI models based on what they can say rather than where they are used—a distinction that could significantly expand government power over the technology sector.
Anthropic had previously discussed the latest models' capabilities with the White House and had limited use of some of its latest technology to select customers because of its ability to surpass human cybersecurity experts in finding and exploiting computer vulnerabilities. The company released Fable, a limited version of the more advanced Mythos model, widely about a week ago.
Why This Matters:
This case illustrates a fundamental tension in technology governance: how democratic societies balance national security concerns with the need for robust public institutions and competitive markets. The government's ability to unilaterally shut down deployed AI systems without court approval or transparent criteria sets a concerning precedent for executive power over private companies. When security experts across the industry warn that the action weakens America's own defensive capabilities while signaling to international partners that U.S. technology companies operate at government discretion, it raises questions about both effective governance and the long-term competitiveness of American technology. The lack of transparency around the government's specific security concerns—acknowledged even by some reporting outlets—suggests that affected companies and the public lack meaningful opportunity to understand or challenge the rationale for such interventions. Whether through regulatory frameworks, court oversight, or mandatory transparency requirements, democratic accountability in AI governance remains underdeveloped relative to the government's demonstrated capacity to intervene.