Anthropic disabled access to its Fable 5 and Mythos 5 artificial intelligence models worldwide on Friday, June 12, 2026, following a U.S. Commerce Department directive. The government cited national security concerns regarding potential model vulnerabilities, forcing the company to restrict access for all users, including foreign nationals both inside and outside the United States.
Commerce Department Export Controls and Worldwide Shutdown
The U.S. government issued an order at 5:21 p.m. ET on Friday that effectively halted the distribution of Anthropic’s most advanced AI technology. According to the Los Angeles Times, the Commerce Department mandate required Anthropic to suspend access to Fable 5 and Mythos 5 for any foreign national, regardless of their physical location. Because the company lacks a mechanism to screen users by nationality in real time, it chose to take both models offline globally to ensure total compliance with the export control directive.
This move marks a significant escalation in the federal government’s oversight of artificial intelligence. While previous administrations limited access to hardware like semiconductors, this directive targets the software itself. As Fortune reports, the restrictions apply even to Anthropic’s own non-citizen employees, complicating internal operations for the San Francisco-based developer.
The Dispute Over “Jailbreak” Vulnerabilities
The government’s decision stems from reports of a security vulnerability in Fable 5. Anthropic stated that officials identified a technique—described as a narrow, non-universal jailbreak—that could allow users to bypass safety guardrails and access cybersecurity capabilities. TechCrunch notes that this specific capability involves prompting the model to analyze codebases for software flaws, a task the company argues is already common among other frontier models, including OpenAI’s GPT-5.5.
Industry experts have questioned the government’s reliance on this research. Katie Moussouris, CEO of the cybersecurity firm Luta Security, reviewed the findings and characterized them not as a malicious jailbreak, but as “Defense Oriented Prompting.” She argued that such capabilities are necessary for security professionals, suggesting that the government’s intervention may have inadvertently hindered defensive research.
For more on this story, see Anthropic halts Fable 5, Mythos 5 after U.S. security directive.
“We disagree that the finding of a narrow potential jailbreak should be cause for recalling a commercial model deployed to hundreds of millions of people. If this standard was applied across the industry, we believe it would essentially halt all new model deployments for all frontier model providers.
Strained Relations with the Federal Government
The shutdown occurs against a backdrop of ongoing friction between Anthropic and the Trump administration. Earlier this year, the Department of Defense designated Anthropic a “supply chain risk” after the company resisted contract terms that would allow its models to be used for autonomous weapons systems or mass domestic surveillance. Yahoo News reported that the administration’s skepticism toward the company has been building for months, with the Pentagon previously barring federal agencies from using Anthropic’s models.
The timing of the order, coming only three days after the public launch of Fable 5, has fueled speculation regarding the administration’s strategic goals. While the government has not provided specific technical details beyond verbal evidence of the jailbreak, the impact on Anthropic’s business is immediate. The company, which recently held a funding round valuing it at $965 million, is preparing for an initial public offering, a process that analysts suggest could be complicated by these regulatory hurdles.
Market Implications and Future Outlook
Industry observers warn that the government’s action sets a precedent that could affect all major AI labs. If the standard for “national security risk” is lowered to include narrow, remediable vulnerabilities, developers may become less transparent about their own testing processes. Ben Murphy, a scholar at the Institute for Progress, noted that the directive is “another step on the balkanization of technology,” potentially driving AI development toward in-house labs or foreign open-weight alternatives.
For now, Anthropic maintains that the outage is a misunderstanding. The company continues to offer its Claude Opus 4.8 model to the public while it works to resolve the export control issues with the Commerce Department. The next 30 days will likely reveal whether the administration intends to pursue a standardized licensing framework for frontier AI or if this remains a targeted effort to curb Anthropic’s reach in the defense sector.
Find more reporting in our Business section.
