The number — 56% of organizations don’t know how fast they could stop an AI system in an incident

Organizations are deploying AI faster than they are preparing to stop it. In ISACA’s survey, 56% of the professionals polled do not know how long it would take their organization to halt an AI system during a security incident, and 39% do not know whether a documented process even exists to shut it down or override it. An AI system running inside one of your processes has to be stoppable quickly, by a named person, following a procedure everyone knows. As long as no one knows how to stop it, its autonomy remains a risk you do not control. Before you let an AI run one of your processes, write its shutdown procedure on a single page (who can stop it, by what action, in how many minutes) and rehearse it once before go-live.

Source: ISACA, “2026 AI Pulse Poll,” May 2026 (3,400+ professionals).