Anthropic Complies with Government AI Recall

Anthropic has announced that it has disabled two of its most advanced AI models. This decision follows an order from the US government on Friday. The government cited national security concerns as the primary reason for the shutdown. As a result, Claude Fable 5 and Claude Mythos 5 are no longer available to users worldwide. The company received the directive at 5:21 pm Eastern Time. Although Anthropic has complied with the request, it has publicly stated its disagreement. The company believes that the government’s assessment of the risk is incorrect.

Mythos is particularly powerful because it can identify security vulnerabilities in software. It has found flaws in major operating systems and browsers during testing. Because of these capabilities, Anthropic kept Mythos restricted to a small group of 50 organizations. This program, called Project Glasswing, included companies like Google, Microsoft, and CrowdStrike. They used it to improve their defense against hackers.

Fable 5 was designed to be a safer version for the general public. It included guardrails that block responses in high-risk areas. It was released only three days before the government order. Tests showed it was the most capable public model available. However, the government’s concern stems from a reported jailbreak of Fable 5. Anthropic argues that this jailbreak is limited and common in other public models. They pointed out that OpenAI’s GPT-5.5 has similar capabilities.

Anthropic remains frustrated because they have always prioritized safety in their development process. The company is planning an IPO later this year. Some people, like Sam Altman, have criticized Anthropic’s marketing strategy. Altman said that telling people the AI is dangerous is just a way to sell more software. Now, it seems the government is taking those warnings very seriously. This event could change how other companies release new AI models.