Control library

Which controls does the register check?

1. GOV-01Agent inventory. Keep a single, current list of every AI agent in use, with its platform, identity, tools and the systems it can reach. Documented by 6 platforms in the register.
2. GOV-02Named accountable owner. Assign every agent a named person who answers for what it does, and record that owner in the inventory. Documented by 8 platforms in the register.
3. GOV-03Documented purpose. Write down what each agent is for, which tasks it may perform and which data it needs, before it gets access. Documented by one platform in the register.
4. GOV-04Least privilege at creation. Grant each new agent only the permissions its documented purpose needs, at the moment its identity is created. Documented by one platform in the register.
5. GOV-05Short-lived credentials. Issue agents credentials that expire within minutes or hours and are scoped to a task, instead of static keys or long-lived tokens. Documented by 6 platforms in the register.
6. GOV-06Per-action policy. Check each agent action against policy at the moment it happens, using the agent, the person behind it, the tool and the data involved. Documented by 6 platforms in the register.
7. GOV-07Human approval for high-risk actions. Route actions above a defined risk threshold to a named human approver, and block them until that person approves. Documented by 2 platforms in the register.
8. GOV-08No self-approval. Prevent any agent from approving its own action, and prevent the person who requested an action from being its only approver. Documented by one platform in the register.
9. GOV-09Policy testing before go-live. Test new or changed policies against realistic or recorded agent activity before they are enforced on live traffic in production. Documented by one platform in the register.
10. GOV-10Tool allow-listing and hiding. Allow each agent to use only the tools on its approved list, and hide every other tool from it entirely. Documented by 2 platforms in the register.
11. GOV-11Revocation and kill switch. Be able to cut off one agent, or every agent, from all tools and data within minutes, without waiting for a deployment. Documented by 2 platforms in the register.
12. GOV-12Decommission at task end. Remove an agent's identity, credentials and access as soon as its task or project ends, and record when that happened. Documented by one platform in the register.
13. GOV-13Per-action audit log tied to a person. Log every agent action with the agent, the person it acted for, the tool, the decision and the time. Documented by 3 platforms in the register.
14. GOV-14SIEM export. Send agent activity and policy decisions to the security team's SIEM in a structured format, close to real time. Documented by 2 platforms in the register.
15. GOV-15Periodic access review. Review each agent's access at least quarterly, with its named owner confirming or removing every permission and tool. Documented by one platform in the register.
16. GOV-16Shadow agent detection. Find agents that were created without registration, including agents connected through personal accounts or tokens outside single sign-on. Documented by 3 platforms in the register.
17. GOV-17MCP server registry. Keep a registry of every MCP server agents may connect to, with its endpoint, owner and environment, and block unlisted servers. Documented by one platform in the register.
18. GOV-18Delegation bounded by the person's permissions. Ensure an agent acting for a person can never do more than that person is allowed to do. Documented by one platform in the register.
19. GOV-19Agent incident response. Maintain a runbook for agent incidents that covers detection, containment, owner notification, evidence capture and recovery, and rehearse it twice a year. Documented by no platform in the register.
20. GOV-20Change management for agent permissions. Treat changes to an agent's permissions, tools or policies as changes that need a ticket, a reviewer and a record. Documented by no platform in the register.
21. GOV-21Third-party agent onboarding. Assess agents supplied by vendors or partners before they connect, covering purpose, data access, credentials and contract terms. Documented by no platform in the register.
22. GOV-22Data access boundaries. Define which data classes each agent may read or write, and enforce those limits at the data source or gateway. Documented by no platform in the register.
23. GOV-23Evidence retention. Keep agent logs, approvals and review records for a defined period that meets legal and contractual requirements. Documented by no platform in the register.
24. GOV-24Board reporting. Report agent governance to the board or its risk committee on a set schedule, using a short page of consistent measures. Documented by no platform in the register.