The US government has suspended access to Fable 5 and Mythos 5 for all foreign users, including Anthropic employees, citing national security concerns. The directive, issued on June 12, 2026, requires Anthropic to disable these models for all customers to ensure compliance. Access to other Anthropic models remains unaffected. The government did not provide specific details about its national security concerns but stated it has become aware of a method to bypass Fable 5's safeguards. Anthropic reviewed a demonstration of this technique and found it involved previously known, minor vulnerabilities. These vulnerabilities are relatively simple and can be discovered by other publicly available models without requiring a bypass.

Anthropic has implemented strong safeguards to reduce the likelihood of Fable being misused for cybersecurity tasks. These safeguards are so robust that some users have complained they are overly broad. In the weeks leading up to Fable 5's launch, Anthropic collaborated with the US government, the UK AISI, and private organizations to conduct extensive red-team testing. These tests showed Fable's safeguards are more effective than those of any previously deployed model. No testers have yet found a universal jailbreak method that can broadly bypass the model's safeguards. Anthropic acknowledges that perfect jailbreak resistance is not currently achievable for any model provider, and all industry safeguards are vulnerable to non-universal jailbreaks.

Anthropic adopted a defense-in-depth strategy with Fable 5, aiming to make jailbreaks either narrow or very expensive to produce, while combining this with thorough monitoring to detect and shut down any successful attacks. This includes a 30-day data retention policy for customers, which carries real costs for the company but allows for research and mitigation of jailbreaks. Anthropic stands by this strategy, stating it reduces risks to levels comparable with existing models. To date, no concerning non-universal jailbreak has been reported that led to harmful results. The potential jailbreaks disclosed are either benign or minor findings that do not provide Mythos-specific advantages.

Source: anthropic