Cybersecurity researchers aren't happy about the guardrails on Anthropic's Fable (2 minute read)

2026-06-11

Anthropic's public Fable model, based on Mythos, blocks many prompts that mention cybersecurity, even simple code review or blog analysis requests. Researchers report frequent shutdowns, keyword-style filtering, and fallback to Claude Opus 4.8. Anthropic instead directs high‑trust users into a Cyber Verification Program with looser limits.