Anthropic spent 1,000 hours running an external red-team bounty before launching Claude Fable 5. The...

A researcher claims he has already bypassed Claude Fable 5’s safety filters, exposing flaws in Anthropic’s guardrail system using multi-step AI prompts.

Anthropic spent 1,000 hours running an external red-team bounty before launching Claude Fable 5. The...