AI governance
AI registry: what it records, who owns it and where to start
A plain guide for risk, IT and compliance leads who have been asked to keep a list of the AI the organization uses. Facts come from NIST's, Microsoft's, Databricks' and IBM's own pages as read on 11 November 2026.
An AI registry is a central record of the AI systems an organization uses, so someone can see what exists and who owns it. The term is used in more than one way. NIST's AI RMF Playbook, under GOVERN 1.6, says mechanisms should be in place to inventory AI systems, and suggests naming a specific individual or team responsible for maintaining the inventory. Microsoft's Agent 365 offers a centralized registry of agents, and Databricks offers a model registry in Unity Catalog. Our view: start with a plain list your team keeps up to date, then choose tooling. Sources read 11 November 2026.
plain answer
AI registry: what is it?
An AI registry is the list an organization keeps of its AI systems. NIST's AI RMF Playbook calls the outcome an inventory: GOVERN 1.6 reads "Mechanisms are in place to inventory AI systems and are resourced according to organizational risk priorities."
NIST suggests policies that define how inventories are created and maintained, and a specific individual or team responsible for maintaining the inventory. It prefers inventorying all models or systems, with high-risk or high-stakes systems as the minimum.
Source: NIST AI Resource Center, AI RMF Playbook, Govern (https://airc.nist.gov/airmf-resources/playbook/govern/), read 11 November 2026. The page shows no date and says AI RMF 1.0 is being updated.
plain answer
What do registries look like in real products?
Microsoft says admins can view all their agents in a single, centralized registry showing agent adoption, activity and health, and that Agent 365 is generally available for the Commercial segment as of 1 May 2026. Its page says Agent 365 works best with Microsoft E5 and needs at least one user with a qualifying licence.
Databricks provides a hosted version of MLflow Model Registry in Unity Catalog. Its page says model versions can carry aliases, that lineage is saved automatically once a model is registered, and that access is granted through Unity Catalog permissions.
These are two different things. One lists agents across an organization. The other tracks versions of models you train and register yourself. Neither page says it lists AI tools that staff use on their own.
Sources: Microsoft Learn, Microsoft Agent 365 overview (https://learn.microsoft.com/en-us/microsoft-agent-365/overview), page dates 2026-08-19 and 2026-10-23; Databricks, Manage model lifecycle in Unity Catalog (https://docs.databricks.com/aws/en/machine-learning/manage-model-lifecycle/), last updated Nov 5, 2026; both read 11 November 2026.
how to
How do you start an AI registry?
Our view: start with a spreadsheet or a shared list before buying anything. For each AI system record the name, what it is used for, who owns it, which data it touches, who approved it and when it was last reviewed. Include tools bought by a department and tools staff use on their own accounts, since those are the ones most often missing.
NIST's playbook lists inventory attributes such as documentation, source code links, incident response plans, data dictionaries and contact information for AI actors. Use those as a prompt for what to add once the basics are in place.
Sources: NIST AI RMF Playbook (https://airc.nist.gov/airmf-resources/playbook/govern/), read 11 November 2026. The starting list is Kriv's view.
security
Who owns the AI registry, and who checks it?
NIST suggests a specific individual or team responsible for maintaining the inventory. Our view: name one owner, set a review date, and make adding a system part of approving it, so the list does not go out of date the week after it is written.
For banks, IBM says the US regulation SR 11-7 required banks to maintain an inventory of models implemented for use, under development for implementation or recently retired. IBM also says the revised guidance, SR-26-2, superseded SR 11-7 as of 17 April 2026, and its page does not say whether SR-26-2 keeps that requirement. Check the current guidance if it applies to you.
Sources: IBM, What is AI governance? (https://www.ibm.com/think/topics/ai-governance), published 10 October 2024 and updated 8 July 2026 per the page, read 11 November 2026.
kriv role
Where Kriv helps
Kriv AI helps teams build the first version of the list: finding the AI already in use, deciding what to record, naming owners and setting a review rhythm. Engagements are scoped and priced after a short discovery call.
Kriv does not sell a registry product and does not replace NIST's, Microsoft's or Databricks' documentation. Kriv has no public case study for a governance engagement, so this page describes a method, not a delivered result.
limits
What this page could not confirm
NIST's page shows no date. We did not read Microsoft's registry documentation beyond the overview, Agent 365 pricing, or any other vendor's registry, and we did not test any product. We did not read the text of SR-26-2, so we cannot say what it requires. We did not read any law that requires an AI registry for a particular organization. Check the current pages before you rely on them.
Sources
Cited sources
- NIST AI Resource Center: AI RMF Playbook, Govern (no date shown), read 2026-11-11
- Microsoft Learn: Microsoft Agent 365 overview (ms.date 2026-08-19, updated 2026-10-23), read 2026-11-11
- Databricks docs: Manage model lifecycle in Unity Catalog (last updated Nov 5, 2026), read 2026-11-11
- IBM: What is AI governance? (published 10 October 2024, updated 8 July 2026), read 2026-11-11
Related resources
Continue exploring
Straight answers
Frequently asked questions about AI registry: what it records, who owns it and where to start
What is an AI registry?
A central record of the AI systems an organization uses. NIST's AI RMF Playbook, GOVERN 1.6, says mechanisms should be in place to inventory AI systems (read 11 November 2026).
Who should maintain an AI registry?
NIST's playbook suggests defining a specific individual or team responsible for maintaining the inventory (read 11 November 2026).
Is a model registry the same as an AI registry?
Not always. Databricks' model registry in Unity Catalog tracks versions of registered models, and Microsoft's Agent 365 registry lists agents (both read 11 November 2026). Our view: a full AI registry also covers tools staff use on their own.
Does every organization need an AI registry by law?
We did not find a general requirement and quote none. IBM says SR 11-7 required banks to keep a model inventory and that SR-26-2 superseded it as of 17 April 2026 (read 11 November 2026). Check the rules that apply to you.
Can Kriv help us build our first AI registry?
Kriv AI helps find the AI already in use, decide what to record and name owners. Scope and price are set after a short discovery call.
Talk to the team that would do the work
Bring your requirements to a working session with the person who'll actually deliver.
Book a Discovery Call