Claude Fable 5 safeguards spark controversy

Anthropic's "safety-first" approach with Fable 5 feels increasingly like a moat disguised as a guardrail, alienating the very research community that helped build the ecosystem.

• –Silently falling back to a weaker model (Opus 4.8) for restricted queries destroys predictability and user trust.
• –Gating advanced capabilities behind "Project Glasswing" creates a centralized, closed ecosystem for AI security that leaves independent researchers in the dark.
• –The broad restriction on "frontier LLM research" seems primarily designed to prevent competitors from using Fable 5 for model distillation rather than mitigating actual bioweapon or cyberattack risks.