Anthropic's Fable 5 safeguards block cybersecurity and biology queries

Anthropic released Claude Fable 5, a Mythos-class model, with broad safeguards that can block basic questions about cybersecurity and biology. Users have found the model may not answer simple prompts about topics such as cancer or security because the classifiers can flag benign requests.

When a reporter tested the system with basic cancer questions, Claude switched from Fable 5 to Opus 4.8 and displayed a pop-up before responding: "Fable 5 has safety measures that flag messages on most cybersecurity or biology topics. They may flag safe, normal content as well.

These measures let us bring you Mythos-level capability in other areas sooner, and we're working to refine them." Anthropic said Fable 5 matches the power of Mythos 5 but added safeguards were necessary to release it publicly after Mythos was held back over cybersecurity concerns and limited to a small cybersecurity project.

anthropic, claude fable, fable 5, mythos 5, mythos-class, opus 4.8, classifiers, safeguards, cybersecurity, biology