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

What happened

One day after Claude Fable 5 launched (June 10, 2026) several cybersecurity researchers publicly complained that Fable refuses even standard defensive questions — exploit explainers, malware analysis, reverse engineering. In high-risk areas (cybersec, biology, chemistry) Fable falls back to Claude Opus 4.8.

Context and impact

Anthropic deliberately set a tighter safety threshold after Mythos incidents. The community reads this as an 'over-refusal' problem: defensive security work becomes harder. Demand grows for Opus 4.8 as the 'practical' model, plus pressure on Anthropic to allow per-key opt-in for vetted security firms.

Details

• Hard safety block, fallback to Opus 4.8
• Affected domains: cybersec, biology, chemistry, distillation
• Community: Twitter, LinkedIn, security research blogs