U.S. orders Fable 5 and Mythos 5 blocked

Anthropic has disabled Fable 5 and Mythos 5 for all its customers after receiving an order from the U.S. government based on national security concerns. The measure also affects foreign nationals, including those inside the country or working for Anthropic.

The company received the directive on June 12, 2026, at 5:21 p.m. Eastern Time. The document does not explain what specific threat prompted the decision. According to Anthropic, the government believes someone has discovered a way to bypass Fable 5's safety barriers, a technique known as a jailbreak.

The problem identified by the government

Anthropic says it has received only verbal evidence of a possible limited jailbreak. In essence, the method involves asking the model to read a codebase and fix security flaws.

The company reviewed the demonstration and says it identified a few previously known, minor vulnerabilities. It also maintains that other publicly available models can do the same without needing this trick, including OpenAI's GPT-5.5.

The difference matters: Anthropic says this is not a universal method capable of unlocking a wide range of dangerous capabilities, but a specific technique for a particular use case.

“We have received no disclosure of a concerning, non-universal potential jailbreak that resulted in harmful output,” Anthropic says.

Anthropic defends its safety barriers

The company says it subjected Fable 5's protections to thousands of hours of testing by its own teams, the U.S. government, the U.K. AI Security Institute, or AISI, and external organizations.

According to Anthropic, those tests showed that the barriers were more effective than those of any previously deployed model. No evaluator reportedly found a universal jailbreak before launch.

The company also acknowledges that no provider can guarantee perfect resistance. That is why it designed a multilayer strategy:

• Reduce the likelihood that the model will be used for dangerous cybersecurity tasks.
• Make potential jailbreaks highly limited or costly to develop.
• Monitor model use to detect and block successful attacks.
• Retain customer data for 30 days to investigate new techniques and fix them.

Anthropic maintains that this approach keeps the risks posed by Fable 5 at levels comparable to those of other models already used in the industry.

What changes for users

For now, access to Fable 5 and Mythos 5 is suspended for all users, not just foreign nationals. Anthropic says it took this step to comply with the legal order, although it considers the measure disproportionate.

Anthropic's other models will not be affected. The company says it is working to restore access and will share more details about the vulnerability within the next 24 hours.

Anthropic also questions the precedent set by the decision. In its view, withdrawing a commercial model used by hundreds of millions of people because of a limited jailbreak could slow the launch of new advanced systems across the industry.

The company does not reject the government's authority to block unsafe deployments. What it calls for is a transparent, clear legal process based on technical evidence. The case puts that tension in the spotlight: models must be useful enough to detect and fix real problems, but also controllable enough to prevent those same capabilities from being used against computer systems.

Now it remains to be seen whether the order stands, what technical details the government presents, and whether other models receive similar treatment. The dispute is not only about two names. It could also define what level of vulnerability is considered acceptable before an AI system is removed from the market.