Anthropic Suspends Fable 5 and Mythos 5 Globally After US Export Control Order

Anthropic has pulled the plug on its two most powerful AI models, Fable 5 and Mythos 5, for every user worldwide after receiving a US government export control directive on June 12 at 5:21pm ET. Citing national security authorities, the order bars access to both models by any foreign national inside or outside the United States, including Anthropic's own foreign-national staff. Because enforcing that restriction is technically infeasible, Anthropic disabled both models for all customers. Claude Opus 4.8 and other Anthropic models remain unaffected, but the company told integrators that new Fable 5 sessions will fall back to the default model, existing sessions will end with an error, and Platform API calls will fail.

Fable 5 is the safeguarded public version of Mythos 5, which is reserved for vetted government cyberdefenders and life sciences partners. Anthropic had only begun rolling Fable 5 out to Pro, Max, and Enterprise customers on June 9, with free access promised through June 22, a window abruptly closed three days in. In a statement, the company said it reviewed a single potential jailbreak demo shared by the government and found only minor, already-known vulnerabilities of the kind routinely discoverable on other public models. Anthropic pushed back hard, arguing that "if this standard was applied across the industry, we believe it would essentially halt all new model deployments for all frontier model providers," and noting that comparable capabilities exist in OpenAI's GPT-5.5.

The directive has drawn international reaction, including from the UK's Minister for AI and Online Safety Kanishka Narayan MP, who flagged disruption to UK customers and pointed to a £1.1 billion domestic AI chip investment as part of the case for technological sovereignty. For organizations that had already built workflows around Fable 5, the takeaway is operational: any automated pipelines, red-team tooling, or research assistants relying on the model now need to be retargeted to Claude Opus 4.8 or an alternative provider, and security teams should audit where those API keys are still in use. Users concerned about how their own traffic and data are being routed through AI platforms, or whether their connections are being masked, can run a quick VPN and proxy detector alongside a privacy checkup to confirm their exposure, while developers hardening public-facing assets in the wake of the disclosed jailbreak class should revisit authentication on AI-backed endpoints with a password strength checker.