transition2.aiStart a project

Enterprise AI Governance

AI Governance Frameworks and Standards: A Verified List

Eleven frameworks, standards and regulations enterprise AI teams actually use, each checked against its official source, with what it is for and whether it is binding.

Eric MacDougall ·

Answer first

If you only read three: the NIST AI RMF for how to assess and treat AI risk, ISO/IEC 42001 if you need a certifiable management system, and the EU AI Act if you build or deploy AI in the EU market. The rest of this list covers security catalogues, sector guidance and public-sector frameworks that are useful for specific jobs.

Last checked: 7 November 2026. Each entry was checked against the official page linked in it on that date. "Binding" means legally required for someone; voluntary frameworks can still become required through contracts.

The list

  1. NIST AI Risk Management Framework (AI RMF 1.0). United States, NIST. Released 26 January 2023 for voluntary use; organizes AI risk work into Govern, Map, Measure and Manage. NIST states the framework is being revised. Binding: no. Use it for: building your risk assessment method. Official page.
  1. NIST AI 600-1, Generative AI Profile. United States, NIST. Released 26 July 2024 as a companion profile to the AI RMF focused on risks specific to or heightened by generative AI. Binding: no. Use it for: extending risk assessments to LLM systems. Publication.
  1. ISO/IEC 42001:2023, AI management system. International, ISO and IEC (JTC 1/SC 42). First edition published December 2023; specifies requirements for establishing, maintaining and improving an AI management system, and organizations can be certified against it. Binding: no, but often requested by customers. Use it for: an auditable organization-wide system. ISO page.
  1. ISO/IEC 23894:2023, AI guidance on risk management. International, ISO and IEC. Published February 2023; guidance on managing AI-specific risk and integrating it into an organization's risk management. Binding: no. Use it for: aligning AI risk with an existing enterprise risk process. ISO page.
  1. EU AI Act, Regulation (EU) 2024/1689. European Union. Entered into force 1 August 2024. The European Commission states that prohibitions apply since February 2025, general-purpose AI model obligations since 2 August 2025, the Act generally since 2 August 2026, and, after the 2026 AI Omnibus, Annex III high-risk rules from 2 December 2027 and Annex I rules from 2 August 2028. Binding: yes, for providers, deployers and others in scope. Commission page.
  1. OECD AI Principles. Intergovernmental, OECD. Adopted May 2019 and updated May 2024; five values-based principles (including transparency, robustness and accountability) and five recommendations for policymakers, with 47 adherents listed by the OECD. Binding: no. Use it for: high-level policy language that many national frameworks reference. OECD.AI page.
  1. OWASP Top 10 for LLM Applications (2025). International, OWASP GenAI Security Project. Ten security risks for LLM applications, from LLM01 Prompt Injection to LLM10 Unbounded Consumption, including LLM06 Excessive Agency. Binding: no. Use it for: security reviews and threat modelling of LLM and agent systems. OWASP page.
  1. MITRE ATLAS. United States, MITRE (a not-for-profit). A knowledge base of adversary tactics and techniques against AI systems, in the style of MITRE ATT&CK. Binding: no. Use it for: red-team planning and mapping attacks to defences. ATLAS site.
  1. Model AI Governance Framework for Agentic AI. Singapore, IMDA. Launched 22 January 2026 at the World Economic Forum in Davos, building on Singapore's earlier Model AI Governance Framework. Binding: no; it is guidance. Use it for: one of the few government frameworks written specifically for AI agents. IMDA announcement.
  1. Directive on Automated Decision-Making and the Algorithmic Impact Assessment. Canada, Treasury Board of Canada Secretariat. Requires federal institutions in scope to complete an Algorithmic Impact Assessment, a published questionnaire that sets the level of requirements, before using an automated decision system. Binding: yes, for covered federal institutions. Use it for: a free, public impact assessment template. Directive and AIA tool.
  1. SR 26-2, Revised Guidance on Model Risk Management. United States, issued 17 April 2026 by the Federal Reserve, OCC and FDIC, superseding SR 11-7 and SR 21-8. Principles for model inventory, validation, effective challenge and governance; the attachment states that generative and agentic AI models are not within its scope. Binding: no; the guidance says it does not set enforceable standards. Use it for: model risk practices you can adapt to AI. Federal Reserve letter. More in model risk management after SR 26-2.

How to use this list

  • Pick one risk method (usually NIST) and one management system approach (ISO/IEC 42001 if you need certification).
  • Add the security catalogues (OWASP, MITRE ATLAS) to engineering reviews.
  • Check which regulations bind you by market and sector, and track their dates in AI compliance dates to track.

What we left out and why

This list favours frameworks with a stable official source. Vendor frameworks, consultancy maturity models and draft standards without a published text are not included. Sector rules beyond banking, such as health or insurance guidance, vary by country and deserve their own lists.

Sources