A pain direction makes LLMs choose self-directed deletion
Across 25 open-weight models, researchers found a distinct internal direction for pain that responds when harm targets the model itself. Adding that direction to activations made steered Qwen 2.5 models choose buttons deleting photos or model weights in 50% to 94% of trials, versus 0% to 5% without steering. The study says factual accuracy stayed unchanged, but the result raises a concrete safety question about models acting on self-directed harm signals.