Anthropic disables Fable 5 and Mythos 5 after US export directive

The U.S. government delivered an export control directive to Anthropic at 5:21pm ET on June 12, ordering suspension of all access to Fable 5 and Mythos 5 by any foreign national, whether inside or outside the United States, including foreign national Anthropic employees. Both models had launched just three days earlier on June 9. Anthropic's statement read: "The net effect of this order is that we must abruptly disable Fable 5 and Mythos 5 for all our customers to ensure compliance. Access to all other Anthropic models will not be affected." (Anthropic blog)

Anthropic said the directive cited a potential jailbreak of Fable 5 -- not Mythos 5. The specific technique, per Anthropic's statement, involves "asking the model to read a specific codebase and fix any software flaws." Anthropic said the government provided only "verbal evidence of a potential narrow, non-universal jailbreak" and that the vulnerabilities the technique produced were "previously known, minor" -- findings Anthropic said are "widely available from other models (including OpenAI's GPT-5.5)" and are "used every day by the defenders who keep systems safe." Anthropic stated: "We have not even received a disclosure of a concerning non-universal potential jailbreak that led to a harmful result." (Anthropic blog; Reuters; Politico)

Anthropic is complying with the legal directive but publicly disputed both the evidence threshold and the legal process. Its statement read: "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." Anthropic also stated the action does not adhere to the principles of a "transparent, fair, clear, and grounded in technical facts" statutory process it has previously called for. (Anthropic blog)

The Guardian describes the move as "a major escalation of US efforts to halt foreign adversaries' AI capabilities." Industry reporting frames the directive as a shift from prior export-control focus on hardware and training tools toward restricting access to trained models themselves. Multiple outlets -- including Reuters, Time, and Politico -- situate the event within broader debates on technological sovereignty and how governments should assess, disclose, and act on model vulnerability findings. (The Guardian; Reuters; Time; Politico)

The immediate technical concern is a "narrow, non-universal jailbreak" -- a technique that bypasses model safeguards in specific circumstances but does not broadly unlock restricted capabilities. Anthropic's launch materials described a defense-in-depth approach to Fable 5 safeguards, including thousands of hours of red-teaming with the U.S. government, the UK AISI, and multiple private third-party organizations, plus a 30-day data retention policy for monitoring and mitigating jailbreaks. Anthropic said it disclosed at launch that "perfect jailbreak resistance is not currently possible for any model provider." (Anthropic blog; Politico)

• •Whether the U.S. government publishes written findings or technical details supporting the directive, and how those align with Anthropic's "verbal evidence" characterization.
• •Responses from other governments and EU institutions about access and technological sovereignty, including any moves to license or host alternative frontier models domestically.
• •Follow-up audits or vulnerability disclosures addressing the specific jailbreak Anthropic references, and any regulatory guidance clarifying criteria for restricting trained model access. (The Guardian; Politico; Reuters)