What is AI Agent Governance?
summary
Defines an agent as code that uses an LLM as its brain and shows how governance must move from postmortem logs to active checks at the moment of action. Describes gateways, sandboxes and preexecution policy checks, cites OWASP’s 10 risks, Microsoft and OpenBox products, an agent registry and Google’s A2A agent card, and proposes cryptographic attestation for immutable logs. Demonstrates real risks — finance agents moving money, prompt injection via inaudible frequencies, shadow AI spawning billions of machine‑to‑machine transactions — and urges runtime governance.
AI-written summary · about
chapters (6)
- 0:00Start
- 1:07How AI agents work
- 3:27Managing AI agents in enterprise systems
- 5:29Guardrails, security and runtime governance
- 10:18Agent-to-gent communication and shadow aI
- 13:00Regulation, compliance and the future of AI governance
related talks
22mWhen AI Agents Start Acting, APIs Become Execution Contracts
22mAccountable AI agents
26mOverview of AI Agent Protocols
39mBuilding Safe AI Agents
46mDemystifying AI Agents
34mAI Agents: Transforming Autonomous Decision-Making
06mAI Agents, Sovereignty & the Next Decade of Innovation
40mAgents in Action: Advancing Open Source AI Missions