OpenAI and Databricks are extending their multi-year partnership, begun in September 2025, to bring OpenAI models and tools to enterprise data governed in Databricks, giving AI systems company-specific context and permissions.

The joint architecture connects ChatGPT, Codex and Databricks Apps to OpenAI models, while Databricks supplies governed data through the Lakehouse, request routing and auditing through Unity Gateway, business context through Genie Ontology, and agent memory and state through Lakebase.

For business users, the Genie plugin lets employees ask questions in ChatGPT; Genie uses approved metrics and definitions from Genie Ontology to find data and run queries in Databricks, then returns results such as charts that respect existing data permissions and link to their sources.

With that business context, ChatGPT can also recommend next steps, identify people to involve, share findings through tools such as Slack or email, and take actions users approve.

For engineers, Unity Gateway enables Codex access while keeping the OpenAI key centrally managed, logging requests, setting budgets, tracking spending by team and selecting which GPT model handles different tasks.

Teams can build and evaluate agents with Agent Bricks or the OpenAI Agents SDK, then deploy them as Databricks Apps; Databricks Model Serving runs GPT models alongside company data under Unity Catalog governance without requiring teams to manage an external API key.

Together, the offerings support three adoption patterns: business users asking questions and taking approved actions through ChatGPT and the Genie plugin, engineers using Codex through Unity Gateway, and teams building custom agents with OpenAI models on Databricks.

Other integrations include OpenAI GPT models on Databricks, Databricks Agent Skills for Codex, a Codex data-science plugin, ChatGPT plugins for data analysis and Neon, and a one-click ChatGPT connection that applies users’ existing permissions.