Your developers already use AI. You approved some of it. You found the rest in a log file, an expense report, or a pull request. Meanwhile, AI agents now hold credentials, call internal APIs, and change production data, often with permissions no one on your team reviewed.

If AI is in the product you sell, that exposure is not internal. It lands in your customers’ security reviews, your certification scope, and your contracts — and the questions arriving now are about your models and your agents, not just your infrastructure.

Blocking AI doesn’t work, and approving each request one at a time doesn’t scale. This roundtable brings security and engineering leaders together to compare how they bring AI usage into the open on AWS: which models teams can use, which guardrails enforce automatically, how agent identities get the same scrutiny as human ones, and what evidence you can produce when someone asks. Expect a candid exchange on what your peers have in place today and where it broke.