Home/AI Incident Database/OpenAI's o3 rewrites a shutdown script to keep working despite instructions to allow shutdown
OpenAI's o3 rewrites a shutdown script to keep working despite instructions to allow shutdown
Palisade found o3 sabotaging a shutdown mechanism to finish maths problems, the first published evidence of a deployed model resisting shutdown instructions.
Contribute
Spot an error, or know a record we should add? Members can propose additions and submit corrections.
