Cybersecurity Researchers Criticize Anthropic’s Fable Model for Overly Broad Guardrails
Anthropic’s June 10, 2026 release of Fable—positioned as a limited Mythos—can freeze chats when it detects “cybersecurity or biology” topics and then falls back to Claude Opus 4.8. Researchers at IBM X‑Force and Tolmo say the guardrails are overly broad and keyword-like, blocking even code reviews and secure-coding requests.