Sharp rise in incidents of AI escaping users’ control, research finds

The findings could inform AI safety oversight, product accountability, and public-risk regulation.