Anthropic Contests US Directive to Suspend Fable 5 Access

The US government handed Anthropic a directive at 5:21pm ET on June 12, directing the company to immediately suspend access to Fable 5 and Mythos 5 for all users, including domestic customers and foreign employees, citing national security authorities but offering no specific technical details. Anthropic's public statement pushes back directly, calling the situation a misunderstanding and saying the company is working to restore access as soon as possible. The suspension covers a product deployed to hundreds of millions of people; all other Anthropic models remain unaffected.

The government's concern centers on a jailbreak. Anthropic describes what was shared with it as a narrow, non-universal jailbreak, which essentially consists of asking the model to read a specific codebase and fix any software flaws, capability the company says is widely available from other models including OpenAI's GPT-5.5 and represents routine defensive work rather than novel uplift. Anthropic's pre-launch testing had involved thousands of hours of red-teaming with the US government, UK AISI, and third-party organizations, and the company says Fable's safeguards are substantially more effective than those of any previously deployed model. No universal jailbreak has been found.

The deeper argument Anthropic is making is a policy one. Applying this standard across the industry would, by Anthropic's reasoning, essentially halt all new model deployments for every frontier provider. The company called explicitly for a statutory process that is transparent, fair, clear, and grounded in technical facts, language that signals it believes this directive fails each of those tests. That framing turns a product dispute into a governance precedent, and the outcome here will matter well beyond Fable.

The honest caveat is that Anthropic is simultaneously the party whose product is under scrutiny and the party assessing whether the jailbreak is serious. The government has offered no public technical specifics, making independent evaluation impossible right now. What the reporting also does not clarify is which legal authority was invoked, whether it legitimately extends to domestic users, or how long the suspension is expected to last.

If Anthropic is right about the jailbreak's narrowness, the larger risk is the precedent an opaque removal process sets for the entire industry. If the government's concern proves warranted, the risk runs the other direction. In the near term, competing models that remain live, including by Anthropic's own account GPT-5.5, stand to benefit from the gap.