U.S. Government Forces Anthropic to Shut Down Fable and Mythos Models Over Alleged Guardrail Bypass
Key Takeaways
- ▸The Commerce Department forced Anthropic to shut down Fable 5 and Mythos 5 via an unilateral export control order without requiring court approval or publishing specific technical justifications
- ▸Security experts argue the alleged guardrail bypass cited by the government is a minor, unfixable quirk that does not warrant export control measures
- ▸Leading cybersecurity researchers warn the export control removes critical AI security capabilities from U.S. defenders and undermines America's credibility internationally
Summary
The U.S. Commerce Department issued an enforcement letter on Friday invoking an obscure export control directive that banned non-Americans, including Anthropic employees, from accessing the company's latest models, Fable 5 and Mythos 5, citing unspecified national security concerns. Anthropic complied by immediately taking both models offline to all customers. The government reportedly justified the action by citing a bypass of the models' safety guardrails, but Anthropic expressed uncertainty about the specific technical justification since the letter provided no details.
New details emerging over the weekend cast significant doubt on the government's reasoning. Cybersecurity researcher Katie Moussouris, who reviewed a paper describing the alleged guardrail bypass, concluded that the behavior "should never have triggered an export control." She explained that the core issue amounts to a minor difference in how users prompt the model—asking it to "review code for security issues" versus "fix this code"—a distinction that affects results minimally. Moussouris and dozens of leading security researchers have called for the Trump administration to revoke the order, arguing that removing advanced AI cybersecurity capabilities from U.S. defenders is dangerous.
The swift, unilateral action—requiring no court approval and with no public disclosure of specific justifications—demonstrates the vulnerability of AI companies to government interference. Observers suggest the enforcement may be retaliatory rather than based on genuine technical concerns, raising broader alarms about the reliability and sovereignty of American AI companies in the global market.
- The incident exposes how quickly the Trump administration can interfere with AI company operations and suggests the move may be retaliatory rather than security-based
