Anthropic restricts dangerous topics in Fable 5 model

Anthropic has released its new frontier AI model, Fable 5, which surpasses previous versions in capabilities but includes significant safeguards. The model refuses to answer queries related to cybersecurity, biology, and chemistry due to concerns about potential misuse by malicious actors. While Fable 5 operates on the same underlying architecture as the more restricted Mythos 5 model, it redirects sensitive queries to an older model and warns users. Anthropic acknowledges these restrictions may occasionally block harmless requests but deems them necessary to prevent the AI from assisting in harmful activities, noting that red-team testing has shown robust defenses against jailbreaking attempts.