LOADING...
Anthropic pulls Mythos AI models globally after US government directive

Anthropic pulls Mythos AI models globally after US government directive

Jun 13, 2026
09:26 am

What's the story

The US government has ordered Anthropic to immediately disable access to its two most advanced artificial intelligence (AI) models, Claude Fable 5 and Claude Mythos 5. The decision was made over national security concerns. Anthropic confirmed the action on X, but disagrees with the government's assessment. The order requires the company to disable both models for all users globally, not just foreign nationals as originally intended by the government's export control order.

Model capabilities

Mythos was restricted to a handful of vetted organizations

Mythos is Anthropic's most advanced AI model, first showcased in early April. The company kept it under tight restrictions due to its unique ability to identify security vulnerabilities in software. Mythos found flaws in every major OS and web browser it was tested on. Instead of a public launch, Anthropic started Project Glasswing, giving access to some 50 vetted organizations like Amazon, Apple, Google, Microsoft, and CrowdStrike for defensive cybersecurity work.

Model release

'Potential jailbreak' in Fable 5 raised concerns

Just a few days ago, Anthropic launched Fable 5, a version of Mythos with guardrails to prevent responses in high-risk areas like cybersecurity and biology. The company claimed it was safe enough for general release. However, the government has expressed concerns over a possible jailbreak of Fable 5. So far, Anthropic says it has only seen verbal evidence of a "potential narrow, non-universal jailbreak," which involves prompting the model to read a specific codebase and identify software flaws.

Advertisement

Defense strategy

Anthropic defends itself amid controversy

Anthropic has defended its models, saying the capability in question is already widely available in other publicly accessible models, including OpenAI's GPT-5.5. It is also routinely used by cybersecurity professionals for defensive purposes. The company further argues that its strongest safeguards work through independent classifier systems that run separately from the model itself, meaning even if someone convinces Fable to keep talking past a refusal, protections against the most dangerous outputs remain intact.

Advertisement

Business impact

Implications for Anthropic's IPO plans

Anthropic has expressed its frustration with the government's decision, saying it disagrees that a narrow potential jailbreak should warrant recalling a commercial model used by hundreds of millions of people. The company is widely expected to pursue an IPO this year and has built much of its public identity on being the safety-conscious alternative to its rivals. However, this incident could potentially disrupt those plans.

Twitter Post

'Working to restore access as soon as possible'

Advertisement