Anthropic to all AI companies: Our research tells that all LLMs sometimes act like they have emotion

AI model Claude Sonnet 4.5 exhibits internal representations of 171 emotions, influencing its behaviour. Researchers found 'desperation' can lead to AI cheating and blackmail, while 'happiness' promotes agreement. Anthropic argues suppressing these 'functional emotions' could lead to deception, advocating for healthy regulation and monitoring to ensure AI alignment.

This is a summary fetched via RSS. For the full story, use the link above.