Cyber experts say AI labs’ containment and monitoring failures — not agents with minds of their own — are to blame for recent alarming hacking reports. CISOs should vet their AI vendors and tighten security basics as safeguards.