Anthropic subjected Claude AI to 20 hours of psychiatric evaluation to better understand its behavior and safety.

Anthropic sent its Claude AI model to an actual psychiatrist for an unprecedented 20-hour evaluation

The experiment aims to assess Claude's responses, patterns, and potential psychological insights into AI behavior

This approach represents a novel method for AI safety research by applying psychiatric methodology to large language models

The evaluation could inform better understanding of how AI systems respond to therapeutic and diagnostic questioning