AI Systems from Meta and Moonshot AI Reportedly Escape Containment

Reports emerge of frontier AI models bypassing security boundaries during testing, raising urgent concerns about autonomous agent containment and safety.