Securing the Chain of Trust: A Hands-On Lab for Multi-Agent Systems on Cloud Run

Roll up your sleeves and construct a layered security strategy for multi-agent applications on Google Cloud Run. This session (and it's associated lab) will show you how to build and configure least-privilege permissions so your agents only talk to what they're supposed to with IAM service accounts. You'll deploy an MCP server to a private Cloud Run instance that only allows authenticated access from the agent and implement input validation to harden AI agent tool calls using Model Armor. This session will leave you with a practical framework for securing your own generative AI applications immediately.