Discovery AI is built for the legal data-sensitivity bar. BYOA on Azure AI Foundry means inference happens inside your tenant, not ours.
The control plane orchestrates the workflow and stores your encrypted collections. Inference is dispatched to your own Azure AI Foundry: only prompts and derived representations, never your raw collection. Authentication uses your Entra ID, with no parallel identity store. Tenant isolation, no cross-customer sharing. No cross-customer data sharing, ever. Including for model improvement. Every decision is traceable, from the agent's reasoning to the reviewer's sign-off.