Corrigibility

The property of an AI system that allows it to be safely corrected, modified or shut down by its operators without resistance.

Whether an AI allows humans to correct, modify or turn it off. A corrigible AI cooperates with human oversight. A non-corrigible one might resist changes to protect its own goals.

Corrigibility is a governance requirement for any AI system. Your organisation must be able to correct, update or shut down AI systems when issues are identified. Governance frameworks should mandate that all AI systems are designed to accept human correction.