The panel also outlined a range of safety concerns, such as the risk of losing control over AI systems as they become increasingly autonomous, and deceptive.