Anthropic blocks all customers' access to Fable 5 and Mythos 5
Anthropic cut off customer access to Fable 5 and Mythos 5 after a U.S. government order citing national security concerns.
Intelligence analysis by GPT-5.4 Mini
Anthropic says it disabled access to its new AI models for all customers to comply with a government directive aimed at foreign nationals, while its other models and Claude remain available. The company disputes the move and says the issue appears tied to a possible jailbreak method.
Anthropic made two new robot-brain tools, then the U.S. government told it to shut them off for some people because of safety worries. Anthropic thinks the warning may be about a way to trick the tool, like finding a crack in a lock.
Analysis
What happened
Anthropic says it disabled all customer access to Fable 5 and Mythos 5 after receiving an order from the U.S. government on Friday, June 12. The company says the government wanted it to suspend access for all foreign nationals, whether they are inside or outside the U.S., including Anthropic employees, and that the move was tied to national security concerns.
Anthropic says the restriction only applies to Fable 5 and Mythos 5. Its other models, along with the Claude chatbot, were not affected.
Why these models matter
Fable 5 is Anthropic’s newly launched public model, introduced on June 9. The company says it was designed to bring many of Mythos’ capabilities to more users. Mythos itself is described as a state-of-the-art cybersecurity model that is only available to Project Glasswing partners.
Anthropic also framed Fable as a major step forward, saying its capabilities exceed any previous model it has launched. The company pointed to internal testing where Fable beat Pokémon FireRed, while Claude failed to beat Pokémon Red, the game it was based on.
The security dispute
Anthropic says it had already added strong safeguards meant to reduce misuse in cybersecurity-related tasks. It also argued that no provider can guarantee perfect resistance to jailbreaks, because every model can be vulnerable in ways specific to that model.
According to Anthropic, the government shared verbal evidence of one possible narrow jailbreak that an unnamed entity allegedly passed to officials. The company says it plans to share more detail within 24 hours, but it disagrees that a possible jailbreak should justify recalling a commercial model.
Anthropic says it supports government powers to stop unsafe deployments, but only through a transparent, fair process grounded in technical facts. It says this action does not meet that standard.
Key points
- Anthropic says it disabled access to Fable 5 and Mythos 5 after a U.S. government order on June 12.
- The government reportedly wanted foreign nationals blocked from using the models, citing national security concerns.
- Anthropic says only Fable 5 and Mythos 5 are affected; Claude and other models remain available.
- The company believes the action may stem from a reported jailbreak method, which it says does not justify a recall.
- Anthropic says it will share more details within 24 hours.
If Anthropic and the government resolve the dispute through a clear process, the company could reopen access with stronger safeguards and better transparency. That could make future AI launches safer without blocking useful tools longer than necessary.
If the security concern proves serious, the models may stay restricted or face broader limits, slowing deployment for customers. The case could also make AI companies more cautious about releasing powerful models, especially ones with cybersecurity capabilities.
