“Some Agent Did It” Isn’t an Answer: Build an AI Agent Audit Trail That Holds Up
From publicly posted calendar Checked yesterday
About this event
Pick an AI agent call from yesterday. Can you identify which agent made it, for which user, what it accessed, and which policy allowed it? Typically, evidence is fragmented across pod logs and tokens, with nothing linking them to a decision. When security traces an incident or an auditor requests proof of control, “some agent did it” is often the only answer. This session closes the series by addressing five audit scenarios: incident tracing, checking regulated data access, quarterly reviews, mandate changes, and EU AI Act compliance (effective December 2027). For each, we’ll extract answers from Lynx’s per-call records and OpenTelemetry traces, exporting them directly to your existing security tools. You’ll leave knowing what an AI agent audit record must capture and how quickly to answer these audit requests. Security engineers and architects who own incident response and evidence, compliance and risk teams preparing for agent audits, and platform leads who have to show their platform is under control. Join us to see how an integrated audit trail gives you full visibility into agent activity, ensuring your platform remains compliant, secure, and ready for regulatory scrutiny.