US government kills Fable 5 and Mythos 5 over a code vulnerability jailbreak
At 5:21pm ET on June 12, Anthropic received an export control directive from the US government. By the evening, Fable 5 and Mythos 5 were dark. Not just for foreign nationals, which the order targeted. For everyone. Anthropic could not isolate foreign users from domestic ones quickly enough, so both models went offline worldwide.
Fable 5 had been live for three days. Anthropic rolled it out free to all Pro, Max, and Enterprise customers on June 9. The model that millions were building on Thursday was gone by Friday night.
The directive cites “national security” authorities. The actual trigger, according to Anthropic’s own statement: someone demonstrated that you could ask Fable 5 to read a codebase and fix software flaws. That is the jailbreak. Asking a coding assistant to scan code for bugs.
Anthropic’s description of the government’s evidence is unusually direct. “To date, the government has only given us verbal evidence of a potential narrow, non-universal jailbreak, which essentially consists of asking the model to read a specific codebase and fix any software flaws.”
The company reviewed the demo and found only minor, already-known bugs. The kind any frontier model can surface without a bypass at all.
Fable 5 is the safeguarded sibling of Mythos 5. Both share the same base model. Fable adds safety filters that block or divert sensitive cybersecurity, biology, and chemistry queries. Mythos 5, unrestricted, goes only to vetted government cyberdefenders and life sciences partners. Now both are gone. New sessions fall back to Opus 4.8. Existing Fable 5 sessions terminate with an error. API requests fail. Integrators were told to migrate.
Anthropic is complying but not going quietly. “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.”
The company pointed out that the capability is widely available elsewhere. OpenAI’s GPT-5.5 does the same thing. Security teams use it daily to hunt vulnerabilities. “If this standard was applied across the industry, we believe it would essentially halt all new model deployments for all frontier model providers.”
The UK’s Minister for AI and Online Safety, Kanishka Narayan MP, confirmed the pause hit UK customers too. He framed it as a case for technological sovereignty and pointed to the UK’s 1.1 billion pound AI chip investment.
This is the first time the US government has emergency-pulled a deployed frontier model. No hearing. No public process. No technical review board. An order at 5:21pm on a Thursday, and the model is gone by midnight. Android Authority compared the directive to the Commerce Department’s 2019 Huawei Entity List placement in terms of speed and unilateral reach.
Every other Claude model remains available. Anthropic says it is working to restore access and will share more details within 24 hours.