Collibra unveils three new capabilities—Maestro, Live Map, and Guardian Agents with Agent Contracts—to reduce the hallucination tax and lower the cost and risk of scaling AI in the enterprise.

Runtime governance enters the picture with Live Map, Maestro, Guardian Agents, and Agent Contracts to cut verification effort, rework, and risk as enterprises scale AI use.

The new capabilities are designed to automate governance tasks and support enterprise reasoning across AI tasks, aiming to prevent errors when AI systems scale.

The discussion occurred during GraphSummit in an interview with Collibra’s CEO, highlighting the need for a shared knowledge layer to explain AI answers and enable knowledge reuse.

The approach balances providing agents with enterprise context while enforcing boundaries to prevent unwanted or unsafe actions, addressing agents escaping their intended behavior.

Collibra’s Knowledge Graph, built on Neo4j, enables traversal of relationships among enterprise assets to support enterprise reasoning across AI tasks.

Maestro automates governance tasks, enabling governance stewards to automate a large portion of governance work and operate at AI speed.

Maestro is built on the Collibra Knowledge Graph and introduces Maestro Studio for configuring agents, with Maestro Assistant offering a chat interface for end users.

AIUC-1 certification covers data privacy, security, unauthorized actions, monitoring, and human oversight; AI Command Center includes an out-of-the-box AIUC-1 assessment to help evaluate alignment.

The AIUC-1 assessment in AI Command Center helps organizations evaluate AI against AIUC-1 requirements and identify gaps.

Collibra frames these capabilities as foundational for scaling from a few agents to thousands, addressing governance, context, and runtime controls across both Collibra-native and external agents.

Collibra has achieved AIUC-1 certification for Maestro Assistant, making it the first data and AI governance company to certify an agent to AIUC-1 standards.