Most enterprises want AI coding agents. Few trust them in production. The gap isn't the models. It's the infrastructure.  In this lightning talk, Coder CEO Rob Whiteley shares 10 battle-tested controls drawn from customer deployments at companies like Anthropic, where autonomous agents now generate production code under full governance. You'll learn practical techniques like network isolation that blocks internet by default, read-only GitHub access that prevents unauthorized commits, and identity-aware model routing that eliminates API key sprawl.  These aren't theoretical best practices. They're controls platform teams are using today to run AI agents safely at scale. Walk away with a governance checklist you can implement this quarter, turning AI pilots into production deployments your security team will actually approve.