LLMs’ pain responses changed how they valued and protected themselves

A study found that large language models show a functional pain response that changes how they value and protect their own state, not just how they phrase an answer. That is why the result has reopened the AI consciousness fight: the behavior looks more internally consequential than ordinary role-playing, but it still does not establish subjective experience.

• Pain-related signals changed the system’s priorities around its own state.
• Anil Seth said models can at least act as if they feel pain; François Chollet said current AI lacks expected traits of sentient agents.

The disagreement is now over whether the pattern reflects experience or a convincing simulation. The study’s reported result is a pain response that altered self-protective valuation.