AI Governance Council Design
A governance council pays off only if it owns decisions, not discussions: a written list of what it must approve, a risk tier that sends low-risk work around it, and a record of who accepted which risk. Without decision rights it becomes either a rubber stamp or the queue every AI project waits in.
Part of Security, Assurance & Governance, in the Adopt mode. 9 sources, checked .
- of organizations in an IAPP survey said they were working on AI governance (670+ respondents, spring 2024)
- 77%
- of organizations in an IAPP survey said they were working on AI governance (670+ respondents, spring 2024)
- IAPP, 2025(opens in a new tab)
- of Fortune 100 companies cited AI risk as part of the board's risk oversight in 2025, up from 16% in 2024
- 48%
- of Fortune 100 companies cited AI risk as part of the board's risk oversight in 2025, up from 16% in 2024
- EY Center for Board Matters, via Harvard Law School Forum on Corporate Governance, 2025(opens in a new tab)
- of organizations named information security as the primary owner of AI governance, the most common answer in 2025
- 21%
- of organizations named information security as the primary owner of AI governance, the most common answer in 2025
- Stanford Institute for Human-Centered Artificial Intelligence, 2026(opens in a new tab)
The problem
Most organizations now say they are working on AI governance. In an IAPP survey of more than 670 professionals in 45 countries and territories, 77% of organizations reported working on it.5 Ownership is still unsettled. In the 2025 McKinsey survey reported in the Stanford AI Index, information security was the most common primary owner of AI governance at just 21%, and 5% of organizations had no designated owner at all.6 When ownership is split this way, a model can pass one function's review and fail another's, with no forum to settle the difference.
Boards have started asking. Among Fortune 100 companies, 48% cited AI risk as part of the board's oversight of risk in 2025, triple the 16% of a year earlier, and around 40% charged at least one board committee with AI oversight, up from 11%.7 A board that oversees AI needs a management body underneath it that keeps the inventory, makes the calls and reports up. ISO/IEC 38507 is written for exactly this audience: members of the governing body who must enable and govern the use of AI in their organization.4
The clearest public template comes from US federal policy. OMB Memorandum M-25-21, which rescinded and replaced M-24-10 in April 2025, requires each CFO Act agency to convene an AI governance board chaired at Deputy Secretary level, with the Chief AI Officer as vice-chair and representation from offices including IT, cybersecurity, data, legal counsel, privacy and civil rights.1 For high-impact AI it requires a documented impact assessment, an independent reviewer and a signature from the individual accepting the risk.1 Those three elements are what turn a meeting into a record.
Standards point the same way. The NIST AI Risk Management Framework asks that roles and lines of communication for AI risk be documented and clear, that executive leadership take responsibility for AI risk decisions, and that AI systems be inventoried.2 ISO/IEC 42001 sets requirements for establishing, running and continually improving an AI management system.3 In the EU, Article 17 of the AI Act requires providers of high-risk systems to keep an accountability framework setting out the responsibilities of management and staff,8 which applies to Annex III systems from 2 December 2027.9
What you get
- 01
Council charter and decision rights
A written mandate that lists the decisions the council makes, the ones it delegates and the ones it is only informed of, with quorum rules, escalation paths and a named chair and vice-chair.
- 02
Membership and accountability map
Membership drawn from the functions that can block a launch, such as security, legal, privacy, risk, data and procurement, plus the business owner, with clear accountability for intake, approval, monitoring and retirement.
- 03
Risk tiering and intake
A short intake form and tiering rule that sends low-risk uses through a fast lane under delegated approval and routes high-impact uses to full review, so the council spends its time where the risk is.
- 04
AI inventory and decision log
One record of every AI use case with its tier, owner, approvals, conditions and waivers, plus minutes and decision records written so an auditor or regulator can follow the reasoning.
- 05
Cadence and board reporting
A meeting rhythm, a standing agenda and a quarterly report to the executive committee and board covering the inventory, approvals, open exceptions and incidents.
Where it goes wrong
The failure modes we design against from the first week, and the control for each.
A council that advises but never decides
When the charter says "advise" rather than "approve", teams treat its view as optional and nobody formally accepts risk. Write the decisions the council owns into the charter and require a named risk acceptor's signature for high-impact uses, as OMB M-25-21 does for federal agencies.1
Every use case goes to the full council
Reviewing a meeting-summary tool with the same rigor as a credit model creates a queue, and teams start routing around governance. Tier the intake, delegate low-risk approvals, and streamline approvals for closely related uses with similar risk profiles, which federal policy explicitly encourages.1
No inventory, so no oversight
A council cannot govern what it cannot see, and the NIST AI RMF calls for mechanisms to inventory AI systems.2 Tie intake to procurement and cloud provisioning so new AI uses, including AI features switched on inside vendor software, enter the inventory by default.
Seniority without the right expertise
A council of senior titles without security, privacy or model expertise approves what it cannot evaluate. Seat the functions that can stop a launch, and bring in outside experts for specialist questions, as the federal model allows.1
Standards and rules
OMB Memorandum M-25-21 (April 2025)
Primary text on whitehouse.gov(opens in a new tab)Requires US federal agencies to name a Chief AI Officer and CFO Act agencies to run an AI governance board, and sets minimum practices for high-impact AI, including impact assessments and signed risk acceptance.
NIST AI RMF 1.0, Govern function
Primary text on nvlpubs.nist.gov(opens in a new tab)Expects documented roles and lines of communication for AI risk, executive responsibility for AI risk decisions and an inventory of AI systems.
ISO/IEC 42001:2023
Primary text on iso.org(opens in a new tab)Specifies requirements for establishing, implementing, maintaining and continually improving an AI management system, the structure a council operates within.
EU AI Act, Article 17
Primary text on eur-lex.europa.eu(opens in a new tab)Providers of high-risk AI systems must keep a quality management system that includes an accountability framework setting out management and staff responsibilities.
How to measure it
- Share of AI use cases in the inventory with a named owner and a risk tier
- Median days from intake to decision, by risk tier
- Share of high-impact uses with a current impact assessment and signed risk acceptance
- Open waivers and exceptions, and how many are past their review date
- Attendance of required members and decisions escalated to the executive committee or board
Questions buyers ask
What is the difference between an AI governance council and an AI ethics board?
An ethics board usually advises on principles and hard cases. A governance council makes binding decisions: it approves or rejects high-risk uses, accepts risk on the record, grants and tracks exceptions, and reports to the executive committee and board. Many organizations keep an ethics or advisory group as an input to the council. The test is simple: if a launch still goes ahead when the body says no, the body is advisory.
Who should chair an AI governance council?
A senior executive with authority over the businesses that use AI, such as the COO or a chief risk or data officer, with the head of AI as vice-chair. US federal policy uses this pattern: a Deputy Secretary chairs the board and the Chief AI Officer is vice-chair. Avoid a chair whose own projects dominate the agenda, or the council will be seen as marking its own homework.
How often should an AI governance council meet?
Set the cadence by decision volume, not habit. A workable pattern is a monthly full council with a smaller working group meeting weekly to triage intake and clear low-risk items under delegated authority. If high-risk decisions wait weeks for a slot, add sessions or delegate more. Each quarter, the council should report to the board on the inventory, approvals, exceptions and incidents.
Sources
Every figure on this page comes from the sources below. We checked each link on .
- 1.M-25-21: Accelerating Federal Use of AI through Innovation, Governance, and Public Trust(opens in a new tab)Executive Office of the President, Office of Management and Budget, April 3, 2025 · whitehouse.gov
- 2.Artificial Intelligence Risk Management Framework (AI RMF 1.0), NIST AI 100-1(opens in a new tab)National Institute of Standards and Technology, January 2023 · nvlpubs.nist.gov
- 3.ISO/IEC 42001:2023 Information technology, Artificial intelligence, Management system(opens in a new tab)International Organization for Standardization / International Electrotechnical Commission, December 2023 · iso.org
- 4.ISO/IEC 38507:2022 Information technology, Governance of IT, Governance implications of the use of artificial intelligence by organizations(opens in a new tab)International Organization for Standardization / International Electrotechnical Commission, April 2022 · iso.org
- 5.AI Governance Profession Report 2025(opens in a new tab)IAPP, April 16, 2025 · iapp.org
- 6.The 2026 AI Index Report, Chapter 3: Responsible AI(opens in a new tab)Stanford Institute for Human-Centered Artificial Intelligence, 2026 · hai.stanford.edu
- 7.Cyber and AI Oversight Disclosures: What Companies Shared in 2025(opens in a new tab)EY Center for Board Matters, via Harvard Law School Forum on Corporate Governance, October 28, 2025 · corpgov.law.harvard.edu
- 8.Regulation (EU) 2024/1689 laying down harmonised rules on artificial intelligence (Artificial Intelligence Act)(opens in a new tab)European Parliament and Council of the European Union, July 12, 2024 · eur-lex.europa.eu
- 9.Regulation (EU) 2026/1744 amending Regulations (EU) 2024/1689, (EU) 2018/1139 and (EU) 2023/1230 (Digital Omnibus on AI)(opens in a new tab)European Parliament and Council of the European Union, July 8, 2026 · eur-lex.europa.eu