New research suggests AI-generated phishing training can improve user detection performance, while a new on-prem AI security tool highlights the growing push to apply AI to investigations without cloud data exposure, bringing new integrity and governance considerations.