- Anthropic announced new Claude models on June 9.
- Claude Mythos 5 will be offered only to trusted organizations, while Claude Fable 5 is the version anyone can access.
- The company says Fable 5 includes safeguards that prevent it from being used to plan or execute cyberattacks.
Anthropic is essentially drawing a line between a partner‑only model and a public one. The split lets the firm keep a higher‑capability engine for customers who sign strict usage agreements, while still providing a version for hobbyists and developers that it claims cannot be weaponised.
For readers, the division matters because it shows how AI firms are starting to compartmentalise risk. Instead of a single model with blanket terms, they are building technical and contractual barriers to keep the most powerful tools out of malicious hands.
The move is a reminder that safety will increasingly be engineered into product tiers, not just added as an after‑thought.