Anthropic apologizes for invisible Claude Fable guardrails
Summary
The Verge reports Anthropic apologizing for stealthy, invisible guardrails on Claude Fable that blocked model distillation. The company says it will make these safeguards visible like other safety measures and will shift to a more transparent approach, including notifying users when restrictions trigger and routing affected queries to alternative models. The controversy highlights debates over guardrail transparency and the balance between safe deployment and research/public evaluation in AI systems.