Securing your entire Claude ecosystem: Top tips, free tools, and Harmonic Security's Approach

Most security teams have a Claude problem they haven't fully scoped yet. Claude.ai is on employee laptops. Claude Code is executing commands in developer terminals. Cowork is autonomously reading, writing, and moving company files. These aren't just data exposure risks: agentic AI takes actions, and those actions can be hard to reverse and harder to audit. In this session, Ed Merrett, Head of Security and AI at Harmonic Security, joins Michael Marriott to map the full Claude attack surface, walk through the riskiest agentic behaviors security teams are seeing in the wild, share free tools you can deploy immediately, and explain why on-device detection that understands intent is the only approach that keeps pace with what these agents actually do. If you're building an AI security policy or actively responding to incidents, this session is 45 minutes well spent.