Collibra announced on 5 October that it had acquired trail ML, a Munich-based AI governance company founded in 2023, to add technology it says can block agent actions that breach governance policies before they occur.
Key points
- Collibra says trail ML’s agents can identify applicable governance controls and assess whether they work.
- The company says trail ML’s runtime technology can enforce Collibra policies where AI agents operate.
- Collibra says it serves more than 700 customers, including 78 of the Fortune 500.
Trail ML adds automated control assessments
Trail ML was founded in 2023 by Anna Spitznagel, Nikolaus Pinger and Sven Hölzel, according to Collibra. Spitznagel is trail ML’s co-founder and chief executive. Collibra describes its own role in the combination as providing context for governance across enterprise data, models, applications and agents, while trail ML supplies technology to automate assessments and governance workflows.
Collibra says trail ML’s agents can examine the context of an AI system, work out which frameworks and controls apply, and assess whether those controls are in place and effective. The company says the technology can also analyse supporting evidence and identify gaps. Changes to that evidence can trigger another assessment, according to Collibra, replacing a single assessment with repeated checks as the material behind it changes.
Collibra says the combined products are intended to help organisations apply requirements under the EU AI Act, ISO 42001 and the NIST AI Risk Management Framework. Those frameworks enter the process when trail ML’s agents determine which requirements apply to a particular AI system, according to the company. Collibra says the subsequent assessment checks the controls against the evidence available for that system.
Collibra policies move to agent runtime
The acquisition also adds a different kind of control to Collibra’s platform. Collibra says trail ML’s runtime technology enforces its policies where AI agents run, blocking actions that violate those policies before the actions take place. That is separate from assessing whether a governance control is present or effective: the runtime capability is intended to act when an agent attempts a prohibited action.
Felix Van de Maele, Collibra’s co-founder and chief executive, said the automation would mean “less manual work assessing controls and maintaining evidence” for governance, risk and compliance teams. Collibra says its aim is to connect those assessments with enforcement as agents interact with enterprise data, systems and tools. Spitznagel said the two companies would combine trail ML’s automation and runtime capabilities with Collibra’s governance technology and customer reach.
Collibra cites more than 700 customers
Collibra puts its client base at more than 700 organisations, 78 of them in the Fortune 500, and says its platform oversees more than 2 billion assets. It describes its platform as the product of eighteen years of ontology engineering. The acquisition is intended to bring trail ML’s assessments and runtime controls into that platform, according to Collibra.
