Claude Fable 5

US government directs Anthropic to suspend Fable 5 and Mythos 5 access globally, citing national security jailbreak concern

The US government issued a directive on June 12, 2026 requiring Anthropic to suspend access to Claude Fable 5 and Claude Mythos 5 — its two most capable models — for all users worldwide. The order invokes national security authorities and claims the government discovered a method to bypass the models' safety guardrails. ## What's new The directive orders Anthropic to "suspend all access to Fable 5 and Mythos 5 by any foreign national, whether inside or outside the United States." To achieve compliance at scale, Anthropic disabled the models for all customers globally — not only foreign nationals — because selectively enforcing nationality restrictions across a deployed consumer product is not technically feasible in real time. The government asserts it found "a method of bypassing, or 'jailbreaking' Fable 5." Anthropic reviewed a demonstration of the claimed exploit and says it showed only "a small number of previously known, minor vulnerabilities." The company is working to restore access as quickly as possible. Access to all other Anthropic models — including Claude Sonnet 4.6, Claude Haiku 4.5, and the rest of the API lineup — is unaffected. Key facts: - Models affected: Claude Fable 5 (`claude-fable-5`) and Claude Mythos 5 (`claude-mythos-5`) - Models unaffected: All other Anthropic models remain fully available - Basis for directive: US national security authorities - Government's claimed finding: A jailbreak method for Fable 5 - Anthropic's assessment: The demonstration showed minor, previously known vulnerabilities - No restoration timeline has been announced ## Context Fable 5 and Mythos 5 launched on June 9, 2026 — three days before this directive. They are Anthropic's most powerful production models, featuring a 1M-token context window, always-on adaptive thinking, and new safety properties including reasoning extraction refusal and a mandatory 30-day data retention policy. Fable 5 is the publicly available flagship; Mythos 5 is restricted to Project Glasswing research partners. This appears to be the first known instance of a US government authority compelling a major AI lab to suspend global access to an already-deployed commercial model on national security grounds. Previous US government actions on AI have largely involved export controls on chips, computing access, and model weights — not directives to pull access to live API services. ## Why it matters Anthropoc's public statement directly contests the government's reasoning: "the finding of a narrow potential jailbreak should not be cause for recalling a commercial model deployed to hundreds of millions of people." This framing sets up a potential conflict between what national security agencies consider acceptable risk thresholds for AI capabilities and what commercial AI labs consider proportionate responses to narrow exploits. The compliance mechanics are also significant. Disabling models globally for all customers — not just the foreign nationals the directive names — shows the difficulty of enforcing nationality-based restrictions on cloud AI services. Anthropic could not technically gate on foreign national status in real time, so the entire customer base bore the cost of compliance. If sustained, this action establishes a precedent: US agencies may have authority to mandate global access suspension for commercial AI models, not just export controls or usage restrictions. That would have significant implications for the AI industry, for international customers who rely on US-hosted frontier models, and for Anthropic specifically as it navigates an ongoing confidential S-1 IPO process with the SEC.