US orders Anthropic to shut off Claude Fable 5 and Mythos 5

In short: The U.S. government ordered Anthropic to immediately disable access to two of its AI models, Claude Fable 5 and Claude Mythos 5.

What happened

The U.S. government issued a directive on Friday ordering Anthropic to shut off access to Claude Fable 5 and Claude Mythos 5. Anthropic said it received the order at 5:21 pm ET and has complied.

Anthropic says the order disables the models for all users worldwide, not just people in other countries. The company said its other AI models are not affected.

Mythos 5 is Anthropic’s most capable model, and the company had already kept it tightly limited. Anthropic said Mythos was unusually good at finding security holes in software (like a very fast auditor that spots weak locks), so it shared the model through a controlled program called Project Glasswing with about 50 approved organizations, including Amazon, Apple, Google, Microsoft, and CrowdStrike.

Fable 5, released three days earlier, was described by Anthropic as a version of Mythos built with extra safety blocks that prevent help in high risk areas like cybersecurity and biology. Benchmark tests from Vals AI reportedly showed it performed very strongly for a public model.

The government framed the action as an export control move, which usually means limiting access for foreign nationals. Anthropic says it believes the real concern is a claimed “jailbreak” of Fable 5, meaning a way to persuade the model to do something it is supposed to refuse.

Why it matters

Many people and businesses rely on top AI models for everyday work, so a sudden shutdown can be disruptive. The case also shows how quickly governments can step in when an AI system is seen as useful for finding software weaknesses, even if the company says similar abilities exist elsewhere.

Source: TechCrunch AI