US government orders Anthropic to suspend Fable 5 and Mythos 5 access over jailbreak concerns
The US government issued an export control directive ordering Anthropic to suspend all access to Fable 5 and Mythos 5 for foreign nationals, citing national security concerns over a reported jailbreak technique, with access cut off at 6:59pm Pacific on June 12, 2026.
Score breakdown
The directive marks a US government intervention that abruptly removed two Anthropic frontier models from all foreign national access worldwide, with Anthropic publicly contesting the national security justification as overstated relative to capabilities already available in other public models.
- 01The US government issued an export control directive suspending all access to Fable 5 and Mythos 5 for any foreign national, including Anthropic's own foreign national employees.
- 02Anthropic received the directive on June 12, 2026 at 5:21pm ET; the letter provided no specific national security details.
- 03The government's stated concern is a jailbreak technique involving asking the model to read a codebase and fix software flaws.
The US government, invoking national security authorities, issued an export control directive to Anthropic on June 12, 2026, ordering the suspension of all access to Fable 5 and Mythos 5 for any foreign national — including foreign national Anthropic employees — regardless of whether they are inside or outside the United States. Anthropic received the directive at 5:21pm ET; the letter did not provide specific details of the national security concern. To comply, Anthropic disabled both models for all customers, while access to all other Anthropic models remained unaffected.
The government's concern centers on a reported jailbreak technique — essentially asking the model to read a specific codebase and fix software flaws — that allegedly enables identification of software vulnerabilities.
Anthropic's statement pushes back on the government's rationale. The government's concern centers on a reported jailbreak technique — essentially asking the model to read a specific codebase and fix software flaws — that allegedly enables identification of software vulnerabilities. Anthropic reviewed a demonstration of the technique and found it surfaced only a small number of previously known, minor vulnerabilities, all of which other publicly available models can also discover without any bypass. Anthropic states that the capability level demonstrated is widely available from other models, citing OpenAI's GPT-5.5 as an example, and is used routinely by security defenders. The company noted it had only received verbal evidence of a potential narrow, non-universal jailbreak and committed to sharing more details within 24 hours.
The post also documents the precise moment access was cut off: a polling script against the Anthropic API confirmed `claude-fable-5` stopped responding at 6:59pm Pacific (9:59pm ET) on June 12, returning a 404 `not_found_error` directing users to use `Opus 4.8` instead.
Key facts
- 01The US government issued an export control directive suspending all access to Fable 5 and Mythos 5 for any foreign national, including Anthropic's own foreign national employees.
- 02Anthropic received the directive on June 12, 2026 at 5:21pm ET; the letter provided no specific national security details.
- 03The government's stated concern is a jailbreak technique involving asking the model to read a codebase and fix software flaws.
- 04Anthropic disputes the severity, stating the demonstrated capability is widely available from other models including OpenAI's GPT-5.5, and is used every day by defenders who keep systems safe.
- 05Access to all other Anthropic models was not affected by the directive.
- 06The `claude-fable-5` API endpoint went offline at 6:59pm Pacific (9:59pm ET), returning a 404 error directing users to `Opus 4.8`.
- 07Anthropic stated it had only received verbal evidence of a potential narrow, non-universal jailbreak and planned to share more details within 24 hours.
Topics
Summary and scoring are generated automatically from the original article. We always link back to the publisher and never republish images or paywalled content. Last processed Jun 13, 2026 · 08:58 UTC. How this works →