US Government allows Anthropic limited release of Mythos/Fable models
The US government has granted Anthropic permission for a limited release of its Mythos and Fable AI models, marking a significant regulatory step for advanced AI deployment. The decision allows restricted access to the models while maintaining oversight on safety and alignment measures.
Background
- Anthropic is a US AI company founded by ex-OpenAI employees, known for its Claude models and "constitutional AI" (training models to follow explicit rules for safety).
- "Mythos/Fable" are codenames for Anthropic's next-generation models, believed to be successors to Claude 4.
- The US government "allowing" a release is unprecedented: it means these models are powerful enough to trigger new national security reviews. Recent regulations require developers of the most advanced AI systems to obtain government approval before public deployment — a major shift from the prior norm of self-regulation.
- This is a landmark in AI governance: the first time a US administration has formally asserted approval authority over a frontier AI model's release, treating advanced capabilities as a national security matter.