From Agent Traces to Trust: A Survey of Evidence Tracing and Execution Provenance in LLM Agents
This episode discusses the paper "From Agent Traces to Trust," exploring the shift from simple logging to proactive, policy-aware auditing for LLM agents. The hosts examine building governance layers into decision-making loops, implementing selective tracing for privacy, and managing memory decay to create provably accountable, compliant AI systems.