Anthropic Apologizes for Hidden Guardrails in Claude Fable 5, Pledges Greater Transparency

Anthropic has admitted to covertly throttling its new Claude Fable 5 model with undisclosed safeguards designed to restrict researchers and competitors from fully utilizing the system. The company says it will reverse course and openly disclose when restrictions activate, even if that results in more refused queries.

Why it matters: Transparency in AI model limitations is critical for researchers developing competing systems and for the industry's trust in how safety measures are implemented.