Autonomous Agents Need Guardrails: Security and Infrastructure with OpenClaw

Autonomous agents are already taking real actions on real cloud infrastructure — provisioning clusters, applying configs, repairing drift. The pattern is familiar: a memory file, a heartbeat check, a loop that runs until the system "looks right." But nobody has agreed on what guardrails belong around that loop, who owns blast radius, or what the verifiable-state layer should look like — and that gap is where outages and security incidents live.

In this 60-minute panel, Adam Gordon Bell (Pulumi) sits down with Dor Serero and Roey Zalta (Microsoft) to work through what breaks when agents touch production infrastructure and the security + infrastructure-as-code patterns that keep them productive instead of dangerous. Live with audience Q&A.