Use cases for agentic AI are exploding by the day. They are all subject to the same risks that IAM practitioners have been fighting for decades, but new patterns have introduced additional new risks. LLMs can share innocuous data like whether your employee prefers a window or aisle seat, but using the exact same protocols, they can also share information about what health, HR, or legal issues that employee has been researching. Sarah will draw on her conversations with Fortune 1000 companies to share valuable insights. We’ll explore established protocols like MCP and A2A as well as some you might not have encountered before like strategies for decentralized identifiers or direct cryptographic signing of HTTP messages.