Collibra Acquires trail ML for Runtime AI Governance

A governance policy can describe what an AI agent is allowed to do. The difficult part begins when that agent reaches a live tool, reads changing evidence, and takes an action before the next review meeting exists on anyone's calendar.

Collibra is closing that distance by acquiring trail ML, a Munich company that automates AI governance assessments and runtime controls. Announced on October 5, 2026, the transaction adds trail's agent-powered workflows to Collibra's enterprise AI control plane. Financial terms, consideration, closing conditions, and an integration timetable were not disclosed.

The deal matters because enterprise AI governance is moving from inventories and approval documents toward continuous decisions inside operating systems. Collibra already organizes context across data, models, applications, and agents; trail adds software intended to determine which controls apply, assess whether they are working, and stop disallowed agent actions where those actions occur.

What Collibra Acquired

trail ML was founded in Munich in 2023 by Anna Spitznagel, Nikolaus Pinger, and Sven Hölzel. Spitznagel serves as co-founder and CEO, Pinger as co-founder and CPO, and Hölzel as co-founder and CTO. The company built software for teams managing Governance, Risk, and Compliance requirements around AI and software systems.

trail's product connects governance requirements with evidence and operating workflows. Its agents can analyze the context surrounding an AI system, identify applicable frameworks and controls, assess control effectiveness, flag gaps, and repeat the assessment when supporting evidence changes. Its runtime layer is designed to enforce policies where agents operate, including blocking an action that conflicts with an approved control.

That last capability explains why Collibra describes the acquisition as a move from assessment toward enforcement. A control can be correct when approved and stale by the time an autonomous system reaches a tool. Continuous reassessment brings governance closer to the speed of the system being governed.

Why trail Fits Collibra's Product Direction

Collibra was founded in Belgium in 2008 and is headquartered in New York and Brussels. The company has expanded from data governance and cataloging into a broader platform spanning data quality, lineage, access governance, unstructured information, and AI controls. Collibra says more than 700 customers use its platform and that it governs more than 2 billion assets.

The acquisition fits a product strategy Collibra has been assembling through software and prior deals. In 2025, the company acquired Raito to strengthen data access governance and Deasy Labs to extend governance across unstructured information. trail adds an automation and enforcement layer for governance work that otherwise depends on people collecting evidence, interpreting requirements, and revisiting controls manually.

The sequence is important. Enterprise agents need trusted context, access boundaries, and usable rules, but those layers create value only when the organization can apply them during operation. trail gives Collibra a mechanism for turning policy and context into repeated assessment and intervention.

The Founders Built Around an Organizational Handoff

Anna Spitznagel, Nikolaus Pinger, and Sven Hölzel met through the Technical University of Munich and the Center for Digital Technology and Management. Their founding thesis came from a familiar organizational split: compliance teams understand regulations, technical teams understand the systems, and both lose time translating evidence, responsibilities, and exceptions between them.

trail addressed that handoff with automation rather than another static repository. The company says its software integrates with existing development and governance workflows, helping enterprises prepare for frameworks including the EU AI Act, ISO 42001, and the NIST AI Risk Management Framework. Its product record also moved into runtime enforcement as enterprises began deploying agents with authority to call tools and change systems.

The company had previously announced a €1.45M pre-seed round led by CapitalT, with NP-Hard Ventures, Mozilla Ventures, Common Magic, and business angels participating. That financing is part of trail's company history, not the value of this acquisition. Collibra did not disclose what it paid.

AI Governance Is Moving Into the Execution Path

Governance products have traditionally helped companies document systems, assign owners, record risks, and prepare evidence for an assessment. Those functions remain necessary, yet autonomous agents introduce an operating problem that a periodic review cannot solve alone. An agent can read external content, call an internal tool, move data, or change a configuration while the approved policy remains unchanged in a separate system.

Collibra has been describing this problem publicly as runtime governance. In a September 2026 interview, co-founder and CEO Felix Van de Maele discussed controls that can evaluate an agent action before it proceeds. trail's acquisition turns that strategy into a more specific product commitment: evidence assessment and policy enforcement should happen repeatedly, close to the point of action.

For Governance, Risk, and Compliance teams, that could reduce manual evidence collection and identify outdated controls sooner. For AI builders, it could make approved boundaries visible inside the systems where models and agents work. The commercial question is whether enterprises can adopt those controls without creating a second layer of operational friction around every useful agent action.

What the Announcement Leaves Open

The official release describes trail as acquired, but it does not provide a purchase price, cash or stock breakdown, earnouts, regulatory conditions, or an expected close date. It also does not explain how trail's products will be packaged, priced, or migrated into Collibra, or how existing trail customer contracts will be handled. Those gaps matter because integration quality will determine whether the acquisition produces one operating control plane or another collection of connected governance modules.

The announcement also stops short of customer proof for the combined platform. Collibra and trail describe a coherent technical direction, but enterprises will still have to test policy accuracy, false-positive rates, evidence quality, latency, auditability, and the behavior of controls across different agent frameworks. Runtime enforcement earns trust through the decisions it gets right and the actions it interrupts for defensible reasons.

What This Signals for Enterprise AI

The acquisition makes a clear category bet. As AI agents receive broader permissions, governance vendors will be judged by their ability to connect organizational intent with system behavior, not only by the completeness of their catalogs or assessment records. Context, evidence, and enforcement are becoming parts of the same operating problem.

Collibra is bringing trail's Munich team and product into a platform already used across large enterprises. The integration now has to preserve the founders' automation thesis while meeting the reliability expectations attached to systems that may approve, flag, or block consequential agent actions. Every policy the combined platform translates into operation will carry the same practical question: can the control keep pace without losing the judgment it was meant to protect?