Anthropic disables Fable 5 and Mythos 5 after US export directive

Anthropic disabled its advanced AI models Fable 5 and Mythos 5 for all customers on June 12 after a U.S. export control directive citing national security concerns, including a potential jailbreak. The move escalates U.S. efforts to restrict foreign access to frontier AI capabilities, leaving non-U.S. governments and organizations without access.

Anthropic disables Fable 5 and Mythos 5 after US export directive Anthropic announced on June 12 that a U.S. export control directive has forced it to suspend access to its most advanced models, Fable 5 and Mythos 5, for foreign nationals and, as a compliance step, to "abruptly disable" those models for all customers, the company wrote on its website. Reuters, The Guardian and Politico report the government provided only "verbal evidence of a potential narrow, non-universal jailbreak," and Anthropic said the letter did not include specific national security details. Anthropic also stated that access to its other models is unaffected. News outlets say the order represents an escalation in U.S. efforts to restrict foreign access to frontier AI capabilities and has left some governments and organizations seeking access scrambling. What happened Anthropic published a statement on June 12 titled "Statement on the US government directive to suspend access to Fable 5 and Mythos 5," saying "The US government, citing national security authorities, has issued an export control directive to suspend all access to Fable 5 and Mythos 5 by any foreign national, whether inside or outside the United States, including foreign national Anthropic employees." The company added that "The net effect of this order is that we must abruptly disable Fable 5 and Mythos 5 for all our customers to ensure compliance," and that "Access to all other Anthropic models will not be affected." Anthropic blog Anthropic told reporters and posted that the government did not provide detailed written evidence of the national security concern and that, in its understanding, "the government believes it has become aware of a method of bypassing, or 'jailbreaking' Fable 5." Reuters, The Guardian and Politico report Anthropic described the government's evidence as "verbal evidence of a potential narrow, non-universal jailbreak." Anthropic blog; Reuters; The Guardian; Politico News outlets including Reuters, The New York Times and Time report the directive prompted Anthropic to cut access worldwide to Fable 5 and Mythos 5, leaving non-U.S. governments, EU institutions and some customers without access. Reporting notes this action follows earlier tensions between Anthropic and parts of the U.S. government over military use and supply-chain restrictions. Reuters; The New York Times; Time Editorial analysis - technical context Industry reporting emphasises that the immediate technical concern cited is a potential "jailbreak," a method that can bypass model safeguards. Industry observers and technical literature commonly treat jailbreaks as vectors that can enable models to produce restricted outputs, expose chains of reasoning, or assist in generating exploit code. Companies producing high-capability models typically run red-team exercises, third-party audits and staged previews; Anthropic's public materials state it conducted red-teaming work and restricted early access to Fable-class and Mythos-class systems prior to broader rollout. Anthropic blog; Politico; reporting context Industry context Industry reporting frames the U.S. directive as a notable shift from prior export-control focus on hardware and training tools toward restricting access to trained models themselves. The Guardian describes the move as "a major escalation of US efforts to halt foreign adversaries' AI capabilities," and multiple outlets situate the event within broader debates on technological sovereignty and national-security controls over advanced AI. This episode intersects with ongoing policy discussions about how to assess model risk, disclose vulnerabilities, and balance commercial deployment with national-security concerns. The Guardian; Reuters; Time For practitioners Industry-pattern observations: Companies and research teams that build or deploy high-capability models should expect rising regulatory scrutiny around guardrails and exploitability. Observers note that export controls focused on model access can materially affect cross-border deployments, procurement, and collaboration, and may drive demand for auditable, explainable safeguards and for onshore or licensed instances of models in regulated environments. These are general patterns drawn from reporting on export controls and past technology restrictions, not claims about Anthropic's internal strategy. Industry reporting What to watch - •Whether the U.S. government publishes written findings or technical details supporting the directive, and how those details align with the "verbal evidence" Anthropic references. - •Responses from other national governments and EU institutions about access and technological sovereignty, including any moves to host or license alternative models domestically. Politico; The Guardian - •Follow-up red-team results, third-party audits, or vulnerability disclosures addressing the narrow jailbreak Anthropic and reporters reference. Anthropic blog; Reuters - •Any regulatory or export-control guidance updates from U.S. agencies clarifying criteria for restricting model access. The reporting base for this summary includes Anthropic's public statement and contemporaneous coverage by Reuters, The Guardian, Politico, The New York Times, Time, Gizmodo and other outlets. Scoring Rationale This story represents a major policy escalation: export controls restricting access to trained models directly affect deployments, procurement, and collaboration. Practitioners should track technical disclosures and regulatory clarifications because they can change access models and compliance requirements. Practice interview problems based on real data 1,500+ SQL & Python problems across 15 industry datasets — the exact type of data you work with. Try 250 free problems /problems