U.S. orders Anthropic to suspend Fable 5 and Mythos 5 over security concerns

The federal government ordered Anthropic to suspend access to Fable 5 and Mythos 5 for any foreign national, inside or outside the United States, and the San Francisco company said it had no choice but to disable both models for all customers. Anthropic said it received the directive on June 12, 2026, at 5:21 p.m. ET, and that the order cited national security authorities.

The company said the government did not spell out the specific concern, but Anthropic’s understanding was that officials believed they had found a way to bypass, or “jailbreak,” Fable 5. Anthropic said it reviewed the technique and found only a small number of previously known, minor vulnerabilities, adding that similar weaknesses could be found in other publicly available models without needing any bypass. Access to Anthropic’s other models was not affected.

The move lands as a policy test case for who gets access to frontier AI and under what rules. Anthropic said it had spent thousands of hours red-teaming Fable 5 with the U.S. government, the UK AISI, private third parties and internal teams before launch, and that no tester found a universal jailbreak. The company also said it required 30-day retention of customer data for Fable 5 as part of a defense-in-depth strategy.

The timing made the order especially sharp. Anthropic launched Claude Fable 5 on June 9, 2026, calling it its most capable model ever made generally available and saying it was state-of-the-art across software engineering, knowledge work, vision and scientific research. The company priced it at $10 per million input tokens and $50 per million output tokens, and said Fable 5 and Mythos 5 could work autonomously for longer than any previous Claude models.

Anthropic had also positioned Mythos 5 as a tightly controlled release, saying it would initially be deployed through Project Glasswing in collaboration with the U.S. government. The company said the model offered the strongest cybersecurity capabilities of any model in the world, but only to a small group of cyberdefenders and infrastructure providers. In Fable 5’s release materials, Anthropic said its safeguards were conservative and sometimes blocked harmless requests, triggering in fewer than 5% of sessions on average, while some higher-risk cybersecurity queries were routed to Claude Opus 4.8 instead.

Anthropic pushed back on the government’s action, saying it disagreed that a narrow potential jailbreak should be enough to recall a commercial model deployed to hundreds of millions of people. It warned that if that standard were applied across the industry, it could effectively halt new model launches from frontier AI providers. For now, the decision turns a technical safety dispute into a national-security precedent, with Washington asserting direct control over commercial AI access and Anthropic trying to restore the models under a tighter regulatory frame.