Collibra Buys trail ML to Automate AI Checks and Enforce Agent Policies

Collibra has acquired Munich-based trail ML, adding technology designed to turn AI governance rules into ongoing checks and controls over agent actions. Announced on October 5, 2026, the deal pairs Collibra’s enterprise governance platform with automation that the company says can assess safeguards and block policy violations before an agent acts. The price was not disclosed.

A young governance company joins Collibra

trail ML was founded in 2023 by Anna Spitznagel, Nikolaus Pinger and Sven Hölzel. Its contribution to the acquisition is agent-powered software for examining evidence, identifying applicable requirements, assessing controls and finding gaps. Collibra supplies the enterprise context used to govern data, models, applications and agents.

Collibra’s rationale is that knowing which rules apply is only the beginning. As AI interacts with company data, systems and tools, organizations also need to determine which safeguards are required, verify that they work and enforce them. The acquisition is intended to automate that work rather than leave governance teams maintaining assessments and evidence manually.

AI governance cannot become another manual process that slows organizations down as AI scales

Felix Van de Maele, Collibra co-founder and CEO

Reopening an assessment when evidence changes

The assessment process starts with the context surrounding an AI system. Collibra says trail ML’s agents can use that context to identify relevant frameworks and controls, then assess whether those controls are present and effective. That goes beyond assembling documents: the stated task includes deciding which requirements apply and whether the safeguards satisfy them.

The company names the EU AI Act, ISO 42001 and the NIST AI Risk Management Framework among the regulations and standards this work can support. It describes the technology as a way to put those requirements into operation, not as a guarantee that an organization meets every obligation.

When supporting evidence changes, assessments can be triggered again. Collibra’s intended result is a continually updated view of governance and compliance, rather than a review that captures only one moment. The trigger is important: the proposed system connects changes in evidence to another assessment of the controls.

The next step is enforcement during operation

The other part of the deal concerns runtime controls—safeguards applied while an AI agent is operating. Collibra says trail ML can enforce its policies where agents run and prevent actions that violate them. This is a different job from assessing a system’s safeguards: it applies the policy at the point of action.

Spitznagel, trail ML’s co-founder and CEO, said combining that automation and enforcement with Collibra’s enterprise context and reach would help organizations move faster with AI safely. The acquisition is announced, but the customer rollout remains unspecified: Collibra did not give a timetable for integrating the capabilities into its platform.