US Curbs Anthropic's Fable 5, Mythos 5 AI Access Globally Over Security; Company Disagrees

Top Summary

• What happened: The US government issued an export control directive, ordering AI company Anthropic to suspend access to its Fable 5 and Mythos 5 models globally, citing national security concerns.
• Why it matters: This unprecedented move impacts the deployment of advanced AI models and highlights the growing tension between AI innovation and perceived security risks.
• What changes: Users outside the US, including foreign national employees, no longer have access to Anthropic's Fable 5 and Mythos 5 models. Access to other Anthropic models remains unaffected.
• Who is affected: Anthropic, its global customer base, and any foreign national previously using Fable 5 or Mythos 5.

US Directs Global Halt on Anthropic AI Models

Earlier this week, Anthropic launched Fable 5, a specialized version of its Mythos AI model, making advanced capabilities more widely available to a broader audience. However, just days later, the US government intervened with a significant directive.

The government ordered Anthropic to suspend access to both Fable 5 and Mythos 5 in all countries outside the US. This decision was explicitly made citing national security concerns.

Anthropic confirmed receiving the export control directive on Friday at 5:21 p.m. ET. The company promptly began disabling the models to comply with the federal order.

Anthropic Complies But Challenges Government's Stance

Despite immediate compliance, Anthropic released a 700-plus word statement expressing its strong disagreement with the government's decision. The company argues the directive lacked specific details regarding the national security concern.

"We received the directive from the government today at 5:21pm (ET). The letter did not provide specific details of its national security concern. Our understanding is that the government believes it has become aware of a method of bypassing, or "jailbreaking" Fable 5," the company stated.

Anthropic is now removing access to Fable 5 and Mythos 5 for all users to ensure compliance with the legal directive. They extended an apology to their customers for the disruption.

Dispute Over "Jailbreak" Vulnerabilities

The core of the government's concern appears to be a method of "jailbreaking" Fable 5. Anthropic reviewed a demonstration of this specific technique, finding it exposed only a small number of previously known, minor software vulnerabilities.

The company asserts these vulnerabilities are relatively simple and can also be discovered using other publicly available models without requiring a bypass. They claim the demonstrated capabilities are already widely available in other leading AI systems, including OpenAI’s GPT-5.5.

"However, we disagree that the finding of a narrow potential jailbreak should be cause for recalling a commercial model deployed to hundreds of millions of people," Anthropic added.

Anthropic's Robust Safeguards and Industry Standards

Anthropic stated it instituted strong safeguards for Fable 5, significantly reducing misuse for tasks like cybersecurity. These safeguards underwent thousands of hours of red-teaming with the US government, the UK AISI, multiple private third-party organizations, and internal teams.

• Tests showed Fable 5's safeguards are substantially more effective than those of any previously deployed model.
• No testers have yet found a universal jailbreak, which broadly bypasses safeguards for a wide range of cyber capabilities.
• Anthropic acknowledges that perfect jailbreak resistance is unlikely for any model, as all industry safeguards are vulnerable to non-universal jailbreaks in specific circumstances.

The company adopted a "defense in depth" strategy, aiming to make jailbreaks either narrow or very expensive to produce. This strategy is combined with thorough monitoring and a policy of 30-day retention of customer data to research and mitigate jailbreaks.

Anthropic argues that applying this standard across the industry would essentially halt all new model deployments for frontier model providers. They believe the government should block unsafe deployments through a "transparent, fair, clear, and grounded in technical facts" statutory process, which they feel this action does not adhere to.

Broader Implications for AI Development

The company believes this incident stems from a misunderstanding. They have only received verbal evidence of a potential narrow, non-universal jailbreak, which essentially involved asking the model to read a specific codebase and fix software flaws.

Anthropic plans to share more details over the next 24 hours. They are actively working to restore access as soon as possible, hoping to resolve the issue.

What to Watch Next

The global AI community will closely monitor Anthropic's efforts to regain worldwide access for its Fable 5 and Mythos 5 models. Further details from Anthropic regarding the alleged jailbreak and its response could clarify the situation. This event may also set a precedent for how governments approach regulating advanced AI deployments based on perceived security vulnerabilities.