Why Anthropic Disabled Its Most Powerful AI Model Days After Launch

The US government ordered Anthropic to revoke foreign access to its advanced AI models, Claude Fable 5 and Mythos 5, citing a potential narrow jailbreak as a national security risk. Anthropic suspended access for all users, arguing the directive was overly broad and that similar capabilities exist in other models like OpenAI’s GPT-5.5, while emphasizing its models’ safeguards are stricter than competitors’." "article": "The US government issued an export control directive on Friday, ordering Anthropic to immediately revoke access for foreign nationals—including its own employees—to its two most powerful AI models, Claude Fable 5 and Mythos 5. In response, Anthropic suspended access for all users to comply, stating the move was a misunderstanding and that it would restore access as soon as possible. The directive stemmed from concerns over a potential jailbreak, a technique to bypass AI safety restrictions. Anthropic clarified the government provided only verbal evidence of a narrow, non-universal jailbreak, involving requests to analyze and fix flaws in a codebase. The company argued such capabilities are already available in other models, including OpenAI’s GPT-5.5, and are routinely used by cybersecurity defenders. Mythos 5, the underlying model for Fable 5, was initially previewed in April but kept restricted due to its power. Anthropic warned it could be exploited to hack critical infrastructure or develop bioweapons, prompting a controlled release program called Project Glasswing with 50 vetted organizations, including Google, Apple, Amazon, Microsoft, and CrowdStrike. These partners used it for defensive cybersecurity work. Anthropic criticized the directive’s scope, stating it would halt all new model deployments if applied industry-wide. The company also disputed the need to disable models deployed to hundreds of millions, arguing its safeguards are stricter than competitors’ and often criticized for being overly restrictive. It plans to share additional details within 24 hours. This latest move reflects ongoing tensions between Anthropic and the US government, including a past standoff with the Trump administration over concerns its technology could be used for mass surveillance or autonomous weapons. The Pentagon had previously cut contracts with the company over these issues.