Standards

The standards your board will be measured against.

What each one is, who uses it, and the questions it should prompt in your boardroom. Each links to the official source.

Voluntary framework

NIST AI Risk Management Framework

US National Institute of Standards and Technology · January 2023

A voluntary framework for managing AI risk across the lifecycle, organised around four functions: Govern, Map, Measure and Manage. It is the most widely cited US reference point for what reasonable AI risk management looks like, and a generative AI profile was added in 2024.

Who it is for

  • Organisations of any size that build, buy or use AI
  • Regulators and courts looking for a benchmark of reasonable practice

Certifiable standard

ISO/IEC 42001

International Organization for Standardization · December 2023

The first international management system standard for AI. Like ISO 27001 for information security, it sets requirements for policies, roles, risk assessment, controls and continual improvement, and an organisation can be independently certified against it.

Who it is for

  • Organisations that want external assurance they can show customers, regulators or acquirers
  • Vendors asked by enterprise buyers to evidence responsible AI

Guidance standard

ISO/IEC 23894

International Organization for Standardization · February 2023

Guidance on applying risk management to AI, building on ISO 31000. It explains how AI changes the familiar risk process: identification, analysis, evaluation and treatment. It is not certifiable on its own.

Who it is for

  • Risk and internal audit teams extending enterprise risk management to AI

Intergovernmental principles

OECD AI Principles

Organisation for Economic Co-operation and Development · 2019, updated 2024

Five values-based principles for trustworthy AI, covering human rights, transparency, robustness, accountability and inclusive growth, with recommendations for governments. They were the first intergovernmental AI standard and underpin the G20 AI principles and many national laws.

Who it is for

  • Boards setting the values their AI policy rests on
  • Companies operating across many jurisdictions that want a common baseline

Voluntary guidance

AI agent incident practices

NIST, NSA and CISA · October 2026

Joint US government guidance setting out practices for detecting, containing and learning from incidents involving AI agents that act on an organisation’s behalf. It extends familiar incident response to systems that can take actions, not only produce outputs.

Who it is for

  • Organisations deploying agents that can access systems, data or money
  • Security and incident response teams
Standards are voluntary unless a law or contract adopts them. Vela does not provide legal advice.Back to Regulations & Standards →