Collibra Acquires Trail ML and Cohere Ships North 2: Two Signals That Enterprise AI Governance Is Now a Product Category

Enterprise AI management took two major steps forward on 5 October 2026. Collibra announced its acquisition of Munich-based AI governance company Trail ML, while Cohere introduced North 2 with tighter controls for agents and automations.

The announcements point to the same pressure facing enterprises: AI systems need to scale without leaving governance, security, spending and permissions behind. Companies are looking for tools that can manage agents across their full lifecycle, from assessing requirements to controlling actions in production.

Collibra adds agent-powered governance

Collibra describes itself as the enterprise AI control plane. Its acquisition of Trail ML is designed to add agent-powered automation, continuous assessment and runtime control to that platform.

Trail ML is a Munich-based AI governance company founded in 2023 by Anna Spitznagel, Nikolaus Pinger and Sven Hoelzel. Its agents can review evidence, determine which requirements apply, assess controls, identify gaps and automate governance workflows across existing tools.

The agents examine the context around an AI system, identify relevant frameworks and controls, and evaluate whether those controls work. When supporting evidence changes, Trail ML's assessments can run again, giving governance teams a way to revisit earlier decisions without starting from scratch.

Trail ML describes its platform around two main use cases: AI governance for scaling AI systems and compliance automation across existing tools. Its agents write to customer systems only after a human approves the action, through a mechanism called Copy-on-Write.

"AI governance cannot become another manual process that slows organisations down as AI scales," said Felix Van de Maele, CEO of Collibra.

Cohere builds a tighter control layer for agents

Cohere announced North 2 on the same day. The platform adds user quotas, rate limits, organisation-wide caps and a redesigned agent harness called Harness, along with more detailed analytics for monitoring activity.

North 2 uses memory to preserve an agent's context across sessions and provides shared libraries for knowledge and assets. Skills are reusable capabilities that agents can call. Organisations can create and share reusable agents and automations, then apply more granular controls to how those automations operate across the organisation.

"Memory lets context persist longer across sub-sessions and agents," said Paul Teyssier, VP, Product AI at Cohere. He also said, "North 2 adds more granular controls over how automations are managed across an organisation."

North Admin sets roles, permissions and controls for model usage, and North 2 integrates with existing identity and access management systems. Its security measures include scanning prompts and responses for personally identifiable information and prompt injection, as well as limiting agent actions with policies.

North 2 can run in the cloud, on premises or in fully air-gapped environments. Cohere positions it as a platform for model portability, agent-level security controls and operation across different deployment settings. The platform holds SOC 2 Type 2, ISO 27001 and ISO 42001 certifications.

Why control and context are becoming central

Enterprise adoption data helps explain the focus on governance and orchestration. In a July 2026 survey, Microsoft led primary-platform choices with 41%, followed by Anthropic's Claude at 28%, LangChain or LangGraph at 10% and OpenAI at 7%.

Major providers such as Microsoft, Anthropic and OpenAI have added spending controls, including monthly consumption limits and workspace defaults. Even so, 21% of enterprises exercise only reactive monitoring of agent spend.

Context creates another problem. In the past six months, 68% of enterprises traced a confidently wrong agent answer to missing or inconsistent business context. North 2's memory and shared libraries address that challenge by keeping context available across sessions, agents and teams.

Only 3% of respondents said none of their deployed agents were genuinely orchestrated workflows. That finding shows how much agent activity already involves connected steps rather than isolated prompts.

Those connected workflows also create risk. Security and permissioning limitations ranked as the top risk of provider-controlled orchestration at 37%, followed by vendor lock-in at 23%. Fifty-three percent of enterprises expect a hybrid control plane by the end of 2026.

That expectation matches the direction of both announcements. Collibra is adding automated assessments and human-approved governance actions, while Cohere is combining memory, reusable capabilities, identity controls, spending limits and deployment options.

What this means for enterprise buyers

For procurement teams evaluating agent platforms, the Collibra and Cohere moves signal that governance is no longer an afterthought or a bolt-on. It is becoming a core product category with dedicated vendors, integrations and budgets.

The practical implication is that buyer evaluation criteria should expand. A platform that scores well on capability and cost but poorly on audit trails, role-based permissions and spend controls is now a higher-risk choice. The suppliers that can demonstrate not only what an agent can do but also what it achieved, under what authority and with what audit trail, will win enterprise trust and enterprise contracts.

What this means for suppliers

For companies selling agentic AI solutions, the bar is rising. A pitch deck full of operational metrics will increasingly lose to a competitor that can show customer impact plus evidence of governance.

The suppliers that invest early in outcome measurement, permissioning and compliance automation will differentiate. Those that do not will find themselves in procurement conversations where they have no answer to the question: how do we control what this agent does?

The Agentic Expo angle

Governance, control and context persistence are the themes that will dominate enterprise agent procurement in 2027. Agentic Expo at Olympia London on 23 to 24 March 2027 will bring together the suppliers building these capabilities and the buyers evaluating them.

If you are selecting agent platforms, governance frameworks or compliance tools, the expo floor is where you can ask vendors the hard questions about audit trails, role-based access and real customer outcomes. The suppliers that have invested in these layers will be the ones worth your time.