Anthropic suspends Fable 5 and Mythos 5 following US government directive

Anthropic has disabled access to Fable 5 and Mythos 5 for all customers following a US government export control directive. The directive, issued on June 12, 2026, requires the suspension of access for all foreign nationals, which necessitates a total shutdown of these specific models to ensure compliance.

Scope of Service Disruption

Access to Fable 5 and Mythos 5 is suspended for all users globally. Access to all other Anthropic models remains unaffected.

Basis of the Government Directive

The US government issued the directive citing national security authorities, though it did not provide specific details regarding the exact nature of the concern. Anthropic reports that the government's action is based on the belief that a method for bypassing or "jailbreaking" Fable 5 has been discovered.

Anthropic's internal review of the government's evidence indicates the following:

• The evidence provided was verbal and concerned a narrow, non-universal jailbreak.
• The specific technique involved asking the model to read a codebase and fix software flaws.
• The capabilities demonstrated by this jailbreak are, according to Anthropic, already widely available in other models, including OpenAI's GPT-5.5, and are commonly used by cybersecurity defenders.
• The vulnerabilities identified were minor and previously known.

Anthropic's Safety and Defense Strategy

Anthropic maintains that Fable 5 was deployed with a "defense in depth" strategy, acknowledging that perfect jailbreak resistance is currently impossible for any model provider. This strategy includes:

• Extensive Red-Teaming: Thousands of hours of testing conducted with the US government, the UK AISI, private third-party organizations, and internal teams prior to launch.
• Strong Safeguards: Implementation of safeguards designed to reduce the likelihood of misuse for cybersecurity tasks, which some users have reported as being overly broad.
• Monitoring and Data Retention: A 30-day customer data retention policy to allow the company to research and mitigate jailbreaks quickly.
• Classification of Jailbreaks: Anthropic distinguishes between "universal jailbreaks" (which broadly bypass safeguards to unblock wide cyber capabilities) and "non-universal jailbreaks" (which elicit specific information in limited circumstances). The company states that no testers have found a universal jailbreak for Fable 5.

Position on Regulatory Process

While complying with the legal directive, Anthropic disagrees with the decision to recall a commercial model based on a narrow potential jailbreak. The company argues that applying this standard across the industry would effectively halt all new model deployments from frontier model providers.

Anthropic asserts that while the government should have the power to block unsafe deployments, such actions should occur through a statutory process that is transparent, fair, clear, and grounded in technical facts—principles they claim this specific action did not follow.