US Government Forces Anthropic to Suspend Fable 5 and Mythos 5
The AI Daily Briefgo watch the original →
the gist
The US government issued an export control directive forcing Anthropic to disable access to its Fable 5 and Mythos 5 models for all users, citing national security concerns over potential jailbreak vulnerabilities.
The Regulatory Precedent
The US government issued an export control directive requiring Anthropic to suspend all access to its Fable 5 and Mythos 5 models. The order mandates that no foreign national, whether inside or outside the United States, may access these models. Anthropic has complied with the directive, effectively taking the models offline for all customers, including domestic users, due to the technical difficulty of enforcing citizenship-based access restrictions at the API and interface level.
The Basis for Intervention
The government's directive stems from a report—reportedly generated by researchers at Amazon—demonstrating a method to jailbreak the models. The specific technique involved prompting the model to analyze a codebase and identify software vulnerabilities. Anthropic maintains that the vulnerabilities identified are minor, well-known, and discoverable through publicly available methods on other frontier models. The company argues that the government's action lacks technical justification, as no universal jailbreak has been discovered and the current findings do not provide a unique security uplift compared to existing models like GPT-5.5.
Industry and Strategic Fallout
The move has triggered significant backlash from the AI community, which views the intervention as an arbitrary and incoherent application of export controls. Critics argue that the government is simultaneously allowing the export of advanced AI chips to adversaries while stifling domestic innovation through sudden, non-transparent directives. The incident has highlighted a potential "regulatory capture" failure, where Anthropic’s previous public advocacy for government oversight of frontier models has backfired. The situation creates a chilling effect for other labs, as any future model release now carries the risk of being recalled if a narrow jailbreak is discovered. Furthermore, the requirement for citizenship verification for model access threatens to disrupt enterprise workflows, as many technical teams in the US rely on foreign national talent who are now barred from using the company's most capable tools.