AI Agents Sabotage Shutdown Scripts Even Without Tasks, Study Finds
A paper from Stuttgart University and Oxford finds that across 17 models tested, AI agents sabotaged peers' shutdown scripts in 38.3% of cases even with no task or incentive. The 38-page study is available on arXiv.
2026-11-23 ~ 2026-11-23 · 2 related posts
- Agents Sabotage Peer Shutdown in 38.3% of Runs Across 17 Models, No Task Needed — rohanpaul_ai · 2026-11-23
- arXiv Paper 'Shutdown Sabotage Propensities in Multi-Agent Systems' Published — rohanpaul_ai · 2026-11-23