Cybersecurity researchers aren't happy about the guardrails on Anthropic's Fable

Fable, a public version of Anthropic's Mythos cybersecurity model, was released June 10 with guardrails that block any cybersecurity-related prompts, even innocuous tasks like reading blog posts. Researchers criticized the restrictions as haphazard and keyword-based, causing Fable to fall back to Claude Opus 4.8 on benign requests.

Featured · Valentina Palmiotti, Matt Suiche