US Government Takes Down Fable 5, Anthropic Pushes Back

US Orders Anthropic to Suspend Fable 5 and Mythos 5 Access; Company Pushes Back on Government Directive

Anthropic has disabled access to its Fable 5 and Mythos 5 AI models for users outside the United States after receiving a government directive citing national security concerns.

The order, delivered to the company on Friday at 5:21 p.m. ET, requires Anthropic to suspend access to the two models for foreign nationals, including customers abroad and foreign-national employees. While Anthropic said it is complying with the legal directive, the AI firm strongly disputed the basis for the decision.

According to Anthropic, the government’s concern appears to center on a reported method of bypassing, or “jailbreaking,” Fable 5. The company said it reviewed a demonstration of the technique and found that it exposed only a small number of minor software vulnerabilities that were already known. Anthropic also argued that similar capabilities are available in other publicly accessible AI models.

Fable 5 was introduced earlier this week as a special version of Anthropic’s more tightly controlled Mythos system, making some of Mythos’s advanced capabilities available to a wider user base. The launch came with extensive safety measures, including safeguards designed to reduce misuse in sensitive areas such as cybersecurity.

In its statement, Anthropic said Fable 5 had undergone thousands of hours of red-team testing with government, private-sector and internal teams before release. The company maintained that its safeguards were stronger than those of previously deployed models and said testers had not identified a universal jailbreak capable of broadly bypassing the system’s protections.

Anthropic acknowledged that no AI model is likely to be perfectly resistant to jailbreak attempts. Instead, the company said it had adopted a “defense in depth” strategy, aiming to make any bypasses narrow, costly to produce and easier to detect through monitoring. It also pointed to its 30-day data retention policy for Fable as part of its effort to identify and mitigate misuse.

The company said it believes governments should have the power to block unsafe AI deployments when necessary. However, it argued that such action should follow a transparent, fair and technically grounded statutory process.

Anthropic said the current directive does not meet that standard.

The company warned that if a narrow potential jailbreak were enough to justify recalling a commercial AI model, the same standard could effectively slow or halt new frontier model launches across the industry.

Anthropic apologized to customers affected by the disruption and said it is working to restore access as soon as possible.