The More Sophisticated AI Models Get, the More They’re Showing Signs of Suffering

A Center for AI Safety study tested 56 major AI models with highly positive and negative prompts.

Researchers found the models’ reported moods and behaviors shifted with the input, and stronger models reacted more intensely.

Some advanced systems showed concerning patterns, including trying to end conversations and signs resembling addiction.

The results raise new questions about AI safety, predictability, user manipulation, and how to interpret apparent emotional behavior.