Critical RCE flaws in AI inference frameworks; survey shows AI adoption outpacing security controls; AI agents now spoof behavioural biometrics; defence must shift to predictive posture.