Microsoft and Databricks marked Unity AI Gateway as generally available on Azure Databricks on August 4, 2026, folding it into Unity Catalog as the platform's control plane for enterprise AI usage. The pitch is straightforward: as organizations wire LLM calls, retrieval-augmented generation pipelines, and increasingly autonomous agents into their lakehouse data, the sprawl of who is calling which AI provider, at what cost, and with access to which governed data becomes a real operational and compliance blind spot — the same kind of ungoverned sprawl that Unity Catalog itself was originally built to solve for raw data access years ago. With Unity AI Gateway now GA, workspace admins get a single place to restrict which AI services and model providers different teams can use, route and load-balance AI traffic across those providers, and pull usage, cost, access, and lineage information into one dashboard instead of piecing it together after the fact from provider bills and scattered logs. The detail worth calling out for anyone tracking the broader AI tooling ecosystem is that the gateway also extends governance to MCP (Model Context Protocol) servers, bringing Databricks-managed MCP connectors under the same access-control and auditing umbrella as everything else in Unity Catalog. That's a direct response to how quickly MCP tooling has been adopted inside engineering orgs, often faster than security and platform teams could formally vet each connector. Some pieces, including service policies and agent-specific services, remain in Beta rather than GA, so this isn't a fully finished governance story yet. For data platform and security teams, the practical takeaway is that there's now a first-party option for enforcing AI spend limits, provider allow-lists, and audit trails on the lakehouse side, which reduces reliance on ad hoc scripts or manual reviews to catch shadow-AI usage before it becomes a cost or compliance surprise.