I was invited as a panelist on this TDL webinar examining the rapidly evolving security risks of AI — agentic systems, prompt injection, data poisoning, deepfakes, and the widening attack surface from integrations with cloud, SaaS, and APIs. The panel's central thesis is one that runs through my own frameworks: AI security is an extension of established cybersecurity principles — least privilege, segmentation, governance, visibility — amplified by speed, autonomy, and scale. The conversation also examines the limits of explainability, the dual-use nature of AI, and where defenders can regain ground through better testing, red teaming, and runtime protection.

Cloud, edge, and fog computing shift where compute happens in IoT and connected systems — and where the trust boundary sits with it. In this talk, I examine the security architecture behind that shift: trusted computing, workload isolation, and device and gateway protection across smart manufacturing, smart energy, and mobile deployments. Moving compute closer to the data source reduces latency, bandwidth, and energy consumption — but the architectural discipline is the same one autonomous and AI systems demand today.