Does the data transmission architecture of AI code review tools create a DLP exposure problem at scale that most security teams aren't accounting for? Trying to understand whether this is a widely recognized problem or something specific to our environment. We've been evaluating AI code review tooling and one thing that keeps coming up in our threat modeling is the raw transmission volume. The standard architecture across most tools works like this: developer writes code, tool scrapes context from open files, raw source payload gets sent to an external inference endpoint, suggestions return. That repeats for every AI code review interaction. At 500 developers generating 100 AI code review interactions per day that's 50,000 daily raw source transmissions to external infrastructure. Each one is a potential interception surface, a DLP exposure point, and an audit event. We're not capturing most of those events in any meaningful way right now.…