Immuta Targets AI Security Bottlenecks with Databricks Integration

The partnership introduces Agentic Data Access, a system that validates user identity to dynamically bound agent sessions. Instead of relying on static service accounts, the tool restricts access to specific tables, rows, or columns based on the user's verified identity at the moment of the request. This approach aims to eliminate the need for pre-provisioning individual accounts while maintaining a granular audit trail.

Immuta CTO Steve Touw argued that legacy security frameworks are reaching a breaking point as organizations attempt to scale AI. By shifting to centralized authorization, the new integration allows firms to automate permissions based on task context rather than static job titles. Complementing these technical controls, the new Comply App allows users to query governance policies in plain English, translating complex metadata into audit-ready reports without requiring manual SQL queries.

Scalability remains a core focus of the update, particularly for large enterprises hitting platform limits. The addition of group-based permission assignments allows Immuta’s engine to calculate complex access combinations behind the scenes. This method bypasses traditional principal-to-object restrictions, ensuring that security policies remain performant even as data volumes and AI agent deployments grow. These features are available to all Databricks customers starting today.