AI is becoming increasingly multi-model, multi-agent, and multi-vendor. Developers are adopting coding agents, while business users are interacting with enterprise data through AI experiences like Genie. Many enterprises are also launching custom agents to automate critical internal workflows. As organizations scale from individual AI applications to fleets of agents connected to models, MCP services, APIs, and enterprise tools, governance challenges expand beyond model access alone. Organizations need visibility, runtime controls, security guardrails, and cost management across their entire AI estate.
Unity AI Gateway is Databricks' governance solution for enterprise AI. Built on the foundation of Unity Catalog, it extends governance beyond data and AI assets to the runtime interactions between models, agents, MCP services, skills, and enterprise tools. Available across AI providers, coding agents, agent frameworks, enterprise applications, and custom AI systems, Unity AI Gateway delivers centralized governance, security controls, cost management, and agent monitoring for enterprise AI.
At Data + AI Summit 2026, we're announcing major new innovations across four areas:
AI costs are increasingly fragmented, making it difficult to understand where token costs are occurring and how to optimize them. Today, we're introducing new capabilities in Unity AI Gateway that help organizations gain visibility into AI usage, control costs, and optimize spend across their AI estate.
"On Udemy's data platform, we route all foundation model traffic through Databricks AI Gateway, giving us a single governance layer for the entire lifecycle. From production agents running on Claude to PII detection pipelines that intelligently balance smaller and larger GPT models for cost efficiency, everything is governed consistently with unified access control and clear cost attribution." — Nathan Sullins, Principal Software Engineer, Udemy
As organizations scale AI, the number of models, agents, MCP servers, tools, and skills quickly multiplies. Yet most enterprises still govern each of these assets separately, if at all, leading to fragmented systems, inconsistent access policies, and limited visibility into what agents do and how AI systems are being used.
Today, we're extending Unity Catalog to govern AI assets beyond models. Databricks-hosted models, external model providers, MCP services, agents, and skills can now be registered, discovered, secured, and audited using the same governance framework organizations already use for data. Administrators gain centralized visibility through permissions, lineage, observability, and cost management, while Unity AI Gateway enforces runtime controls across model calls, tool invocations, and agent workflows. Together, Unity Catalog and Unity AI Gateway provide a unified governance layer for both AI assets and AI interactions.
“AI has the potential to enhance efficiency across the enterprise, but scaling AI responsibly requires strong governance, transparency, and trust. Databricks has helped our team build the unified foundation we need to help govern AI systems, protect sensitive data, and support enterprise-wide AI adoption in a regulated industry.” — George Torres, Senior Director of AI Engineering, First American
Organizations increasingly rely on models from multiple providers. Admins can now govern foundation model access through Unity Catalog using the same fine-grained access policies they use for data. Policies can be dynamically applied based on attributes such as model provider, country of origin, approval status, or any governed tag, making it easier to consistently enforce controls as model inventories grow.
As MCP adoption grows, organizations need a consistent way to govern access to enterprise tools. Databricks now provides managed MCP services for applications including Google Drive, Jira, Confluence, Slack, GitHub, and SharePoint, giving teams governed, ready-to-use integrations without managing their own infrastructure.
Now you can also register custom MCP services, creating a centralized inventory of approved tools and integrations. Admins can manage access, enable or disable individual tools, and audit usage, while developers can discover approved MCP services and use them from Databricks, coding agents, Genie, or external agent frameworks.
"Genie was one of the first MCP connections we added to Asana AI Teammates. This integration allows anyone to self-serve answers to complex data questions. Genie has let our human-agent teams get answers faster, in a single shared conversation, by bringing data context from Unity Catalog alongside business context in the Asana Work Graph. With AI Teammates' multiplayer design and enterprise-grade governance, Genie conversations over MCP respect the permissions of the user asking. Customers can take this further by using Unity AI Gateway to enforce additional governance policies and guardrails." — James Davidheiser, Data Architect, Data Infrastructure, Asana
Traditional governance controls determine who can access a model or tool. As AI agents become more capable, organizations also need controls over what those systems can do during a specific interaction.
Today, we're launching Contextual Service Policies in Beta. Now you can allow, deny, or require approval for actions such as modifying files, pushing code, accessing enterprise systems, or interacting with sensitive information.
Policies can be applied based on the user, agent, model, MCP service, tool being invoked, or the contents of the request and response. You can also enforce AI guardrails to mitigate risks, including PII exposure, prompt injection, jailbreaks, unsafe content, and other policy violations.
For example, administrators can require approval before a coding agent pushes code to GitHub, restrict writes to sensitive Google Drive folders, or block requests and responses that contain regulated data.
Unity Catalog now also provides a governed inventory of reusable skills. Teams can register agent endpoints, publish approved skills, and make these capabilities discoverable across your organization through a common catalog experience.
As AI workflows span multiple models, agents, and MCPs, it becomes increasingly difficult to understand what happened during a specific interaction. Teams often need to piece together logs from multiple systems to troubleshoot failures, investigate incidents, or audit AI activity.
Today, we're introducing new capabilities for unified agent monitoring:
“As we continue to expand AI-powered experiences for our users, governance, privacy, and trust remain foundational. We're excited about the potential for Unity AI Gateway to provide a more streamlined approach to governing AI interactions, applying consistent controls, and helping our teams scale AI innovation responsibly.” — Vladislav Nedosekin, AI Director, Flo Health
Unity AI Gateway is built as an open governance platform for enterprise AI. Today, we're expanding our AI governance ecosystem with integrations across AI security, identity governance, data protection, and threat detection, helping organizations bring trusted controls into runtime AI workflows.
Upcoming integrations will extend support to AI security providers, including Alice, CrowdStrike, Cyera, HiddenLayer, Netskope, Noma Security, Obsidian Security, Openlayer, Palo Alto Networks, and Zscaler, as well as identity providers, including Okta, Ping Identity, and Saviynt.
"As enterprises move AI into production, they need a consistent way to govern and secure AI interactions across models, agents, and tools. By integrating Falcon AI Detection and Response with Databricks Unity AI Gateway, CrowdStrike makes the Falcon platform the security layer for AI, delivering the visibility, detection, and protection organizations need to scale AI across the enterprise." —Daniel Bernard, Chief Business Officer, CrowdStrike
AI agents need an identity, explicit guardrails, and real accountability. Together with Databricks, we are solving this challenge. By connecting Okta’s Identity Platform with the Unity AI Gateway, enterprises can extend consistent governance and trusted controls to agents and the critical data pipelines they touch — Harish Peri, SVP & GM, Okta for AI Agents
“As enterprises scale agentic AI, security needs to move into the runtime path. Together with Databricks, Prisma AIRS will help customers inspect AI interactions in real time and enforce protections across models, applications, and agents.” — Ian Swanson, VP, Product, AI Security at Palo Alto Networks
"AI is quickly becoming a core part of enterprise operations, making governance and security foundational requirements for successful adoption. By combining Databricks Unity Gateway capabilities with Zscaler’s AI Guardrails, customers receive greater visibility, intent-based and outcome-based controls, and protection across AI interactions." — Ashwin Kesireddy, VP, Product Management - AI Security, Zscaler
Omnigent was just announced as an open source meta-harness for building and running agents across models, frameworks, and coding tools. Today, we’re introducing the managed version, Omnigent on Databricks, in Beta. It’s the same Omnigent you run in open source, so there’s nothing to rebuild. You bring your existing setup, harnesses, workflows, and skills, and deploy them to Databricks to run as managed workflows with shared history, remote access, collaboration, and isolated cloud execution on Lakebox.
Unity AI Gateway governs every Omnigent interaction with centrally defined policies, cost controls, smart routing and unified telemetry, giving teams a managed way to run agent workflows alongside data in Unity Catalog. Learn more about Omnigent by checking out our blog.
Enterprise AI is becoming increasingly open, distributed, and agent-driven. With these new capabilities, Unity AI Gateway provides a unified governance layer for models, agents, MCP services, and skills, combining runtime controls, agent monitoring, spend management, and ecosystem integrations in a single platform.
Get started with Unity AI Gateway by checking out our documentation and blog series.
Learn more about our vision for unified AI governance by watching our Agent Platform keynote.
Register for Data + AI Summit and explore our AI governance and Unity AI Gateway product sessions.
此内容由惯性聚合(RSS阅读器)自动聚合整理，仅供阅读参考。 原文来自 — 版权归原作者所有。