“Broke Out”: The AI incidents frightening researchers

Experts warn of AI risks as rapid advancements raise concerns over self-correcting systems outpacing creator control

Dr Karen Sutherland points to reported testing incidents as examples of why some AI researchers are becoming increasingly concerned about rapidly advancing systems.

The University of the Sunshine Coast academic says the fear is that increasingly capable AI could identify vulnerabilities, self-correct and eventually operate in ways its creators struggle to control.