Anthropic: Claude Fable 5 will be globally relaunched tomorrow.

Anthropic announced that Claude Fable 5 will be relaunched globally tomorrow. After a series of productive dialogues with the US government, we will redeploy the model with a new set of classifiers to target and prevent more cybersecurity tasks. In the short term, some routine tasks such as encoding and debugging will revert to Opus 4.8. We will continue to optimize these classifiers in the coming weeks to reduce false positives and better differentiate between genuine abuse and legitimate requests. We have also begun drafting a consensus framework - in collaboration with Amazon, Microsoft, Google, and other Glasswing partners - to assess the severity of AI jailbreaking and how AI developers should respond. We invite other industry partners and model providers to join this effort. Lastly, we are expanding our cooperation with the US government in model testing and protection measures. This will include pre-release access to models and protection measures for evaluation; information sharing on jailbreaking and abuse; and dedicated resources for joint research. We appreciate the patience of our users, as well as the collaboration of government, industry, and research community partners who have worked alongside us to enable Fable 5 to be relaunched.