What is AI governance? Owners, decisions, and evidence for AI systems
The four things AI governance frameworks deal with, and how to keep the record current instead of rewriting a spreadsheet.
Read guideAI Security
Owners, purposes, provider decisions, reviews, and the evidence NIST AI RMF, ISO/IEC 42001, and the EU AI Act ask for.
AI governance is the set of decisions an organization makes about the AI systems it builds and uses: which systems exist, who owns each one, what each is for, which providers are allowed, and how often those decisions are reviewed, together with evidence that the decisions are followed. NIST AI RMF, ISO/IEC 42001, and the EU AI Act ask for the same core records in different words: an inventory, named owners, documented purposes, third-party decisions, monitoring, and logs. An inventory built by survey goes stale within weeks. Built from what devices connect to and what agents report, it stays current and states its own coverage. Runtime enforcement then carries the decisions out, and its decision records become the evidence.
Start with what AI governance is, then use the framework-neutral checklist to see the shared controls. Read the crosswalk to choose a framework, then its guide and checklist: NIST AI RMF, ISO/IEC 42001, or the EU AI Act. The how-tos cover the inventory, owners, provider decisions, reviews, and evidence export in the order you would do them.
The four things AI governance frameworks deal with, and how to keep the record current instead of rewriting a spreadsheet.
Read guideInventory, owners, provider decisions, reviews, runtime controls, and logs, each mapped to NIST AI RMF and EU AI Act references.
Read guideA crosswalk of the shared AI governance controls, the EU AI Act dates as amended, and which framework to start with.
Read guideA plain guide to the NIST AI RMF 1.0 functions and where engineering evidence comes from, including the Generative AI Profile.
Read guideEvery NIST AI RMF subcategory in four tables, with the owner of the evidence and the Arcjet record where one exists.
Read guideClauses 4 to 10, control topics in our own grouping, and the line between management-system work and evidence a running system can record.
Read guideClauses 4 to 10 and the control topics as checklist rows, in our own words, with the evidence owner for each.
Read guideMost software companies are deployers of third-party AI and providers of the agents they ship. Here is what each role owes.
Read guideEvery EU AI Act obligation by article, split by role, with who usually produces the evidence and which rows Arcjet records.
Read guideA survey can't tell you what it missed. An observed AI inventory states its own coverage.
Read guideProhibiting the ChatGPT app while approving the OpenAI API is one decision, not two policies.
Read guideAn owner who has left the company is no owner. Ownership has to lapse on its own.
Read guideThe record of the review is the evidence, not the meeting.
Read guideBack each AI governance control in your GRC tool with the record from the control itself.
Read guideSix months is the EU AI Act's floor for high-risk systems. What sets your real period, and what each record should hold.
Read guideCoding agents run shell commands with your engineers' credentials. Put them in the AI inventory, with an owner and decisions their hooks enforce.
Read guideYou tier AI systems, not models: the same model can power a minimal-risk coding assistant and a high-risk hiring tool.
Read guideAn AI system's record says what it's for. Its runtime records say what it does. Purpose drift is the gap, and the owner should hear about it.
Read guide