Anthropic disables Fable 5 and Mythos 5 after U.S. export control order

Anthropic has disabled access to Claude Fable 5 and Claude Mythos 5 following a U.S. government export control directive, the company said in an official statement.

The order requires Anthropic to block foreign nationals from accessing both models, regardless of where they are located. The restriction also applies to foreign employees inside Anthropic itself. To comply with the directive, the company said it had no practical option but to disable Fable 5 and Mythos 5 for its entire user base. Access to Anthropic’s other models is not affected.

A notice about the restriction is now also being shown inside Claude apps:

According to Anthropic, U.S. officials acted after learning about a jailbreak affecting Fable 5. The company reviewed the method and concluded that it exposed only minor, previously known vulnerabilities, adding that similar results could be achieved with publicly available models, including OpenAI's GPT-5.5. Anthropic said the jailbreak essentially involved asking the model to review a codebase and fix software bugs — a routine task for cybersecurity professionals.

U.S. officials disagree. David Sacks, co-chair of the White House's science and technology advisory council, said a trusted testing partner discovered a jailbreak capable of bypassing Fable 5's safeguards. According to Sacks, the administration asked Anthropic CEO Dario Amodei either to address the issue or stop distributing the model. When the company refused, export restrictions were imposed.

The Wall Street Journal later reported that the discussions involved Amazon CEO Andy Jassy and senior U.S. officials, including Treasury Secretary Scott Bessent. Jassy reportedly told officials that Amazon researchers had used a series of prompts to get Fable 5 to provide information that could be useful in preparing cyberattacks and should have remained inaccessible.

Amazon said the U.S. government regularly consults the company on information security matters. When such exchanges take place, the company said it does not disclose their contents.

Anthropic said it is complying with the regulator’s order but does not agree with the reasoning behind it. In the company’s view, the discovery of a narrow and unstable jailbreak should not be enough to force the withdrawal of a commercial model used by a broad user base.

The company also warned that applying such a standard across the industry could effectively freeze the release of new AI models. If every limited vulnerability becomes grounds for pulling a model from the market, Anthropic argued, the precedent could reshape how advanced AI systems are developed and distributed.

The restriction comes shortly after Anthropic introduced Claude Fable 5 and Claude Mythos 5 as its new fifth-generation models. Fable 5 was positioned as the public version with stricter safety controls, while Mythos 5 was reserved for a limited group of partners and cybersecurity use cases.

Earlier, Durov’s Code covered the launch of Fable 5 and Mythos 5 and explained what made the models different:

Anthropic launches Claude Fable 5, the first model in the new Mythos class

Anthropic has unveiled two new fifth-generation AI models, Claude Fable 5 and Claude Mythos 5.

Durov˙s CodeIgor Savkin