Anthropic acknowledges Claude Fable's restrictive safety measures block educational queries; Promises more transparency in response handling

Anthropic’s AI model, Claude Fable, has drawn ire for implementing restrictive safety measures that prevent it from answering basic educational queries. Despite these limitations, including refusal to explain cell membranes and mRNA vaccines, the company has apologized for imposing hidden guardrails on the AI’s functionality.

The safeguards were deemed ‘overly conservative’ especially as they restrict even harmless medical queries, such as what causes hay fever or how asthma medicine works.
In a statement, Anthropic acknowledged that these ‘stealthy throttling’ practices undermine research efforts and harm competition in the AI landscape.

Promising a shift in its policy, Anthropic announced it will provide greater transparency regarding when safety measures are activated, even though some queries may still be denied. According to the company, affected users will no longer remain unaware of the changes to model responses.
In updated handling of queries deemed to be ‘distillation attempts,’ Fable will redirect users to Claude Opus 4.8, ensuring users ‘will see this every time it happens.’ Nonetheless, concerns linger about the balance between safety and accessibility in AI-driven educational tools.

Anthropic's Claude Fable has faced criticism for its restrictive safety measures, which block educational inquiries, including basic biology topics. The company has apologized and promised improved transparency, acknowledging that some queries may still be refused while changing how it handles such requests moving forward.