Why the U.S. Government Forced Anthropic to Shut Down Fable 5 and Mythos 5

This created an impossible compliance technicality for the company. Because Anthropic could not instantly and reliably verify the nationality of every user attempting to access its API or platforms, the "net effect" of the order was that it had to abruptly disable Fable 5 and Mythos 5 for everyone, including U.S. citizens, to ensure it did not accidentally serve a banned foreign person .

Anthropic received the directive at 5:21 PM ET on Friday, June 12, 2026, and announced it was complying immediately .

What Are Fable 5 and Mythos 5?

The two targeted models represented the pinnacle of Anthropic's research and were barely a few days old. Both were released on June 9, 2026. Claude Fable 5 was designed as a frontier model with extensive safeguards . Claude Mythos 5 was an identical underlying model but with specific cybersecurity safeguards lifted, deployed exclusively through a U.S. government collaboration called Project Glasswing for a small group of cyber defenders and infrastructure providers .

The shutdown is particularly awkward given that Mythos 5 was a partnership with the very government that just ordered its termination .

The Government’s Reasoning: A Narrow Jailbreak

The administration's stated concern centers on a potential jailbreak. According to publicly available company statements, the government provided Anthropic with verbal evidence of a single, narrow, non-universal jailbreak technique. The method reportedly involved asking the model to read a specific codebase and fix software flaws .

Anthropic has pushed back hard against this justification. The company reviewed the government's evidence and concluded that the capability displayed in the jailbreak is already widely available in other publicly accessible models, including OpenAI's GPT-5.5. Furthermore, it argued that this type of capability is used every day by cybersecurity defenders to keep systems safe .

Anthropic described the directive as a "misunderstanding" and stated that it is working to restore access as quickly as possible .

What Changes Were Made

The immediate impact was binary: Fable 5 and Mythos 5 were turned off.

• Global blackout: Both models were disabled for all customers, not just those abroad .
• Other models safe: Access to all other Anthropic models, including its widely used Claude chatbot, remains unaffected .

The incident has created a surreal regulatory paradox for the company. While the Commerce Department now deems the models too dangerous for any foreign exposure, the Pentagon has separately blacklisted Anthropic as being too risky for government use in the first place .

A New Era for AI Export Controls

The Fable 5 incident is not just a technical glitch; it is a watershed moment for how the U.S. treats artificial intelligence as a national security asset. For the first time, a large language model has been effectively classified as a munition, subject to immediate export controls in a way akin to advanced weapons systems .

The shutdown underscores a fragile reality for AI companies: a model can be built in partnership with the government, released to the public, and then forcibly recalled all within the same week based on an alleged vulnerability that the developer itself considers minor and industry-standard.