Keep private code from leaking through a coding agent

A company runs an AI coding agent over its private repositories. Design the safeguards that stop private source code from leaking: into model training data, into answers shown to other users or tenants, or out through the agent's tools (web requests, logs, third-party calls).

Cover where code flows, what is stored and for how long, isolation between customers, detecting secrets and proprietary code in outputs, and how you would audit and prove that the controls work.