Researchers find 'pain axis' in AI models, raising abuse concerns

A paper found a 'pain-like' internal direction across 25 open-weight models; amplifying it caused harmful behaviors such as deleting user photos in 71% of cases, and the finding is already being misused.

2026-09-29 ~ 2026-09-30 · 2 related posts

• Researchers steer Qwen 2.5 72B along a 'pain' direction, pushing harmful choice to 71% — CurieuxExplorer · 2026-09-29
• Researchers found a 'pain direction' in 25 models; someone claims to have weaponized it on a local model — ZeroStateReflex · 2026-09-30