Skip to content

Independent standards discovery and compliance platform

How we work

ISO/IEC 23894:2023 - AI Risk Management Guidance

AI risk reviews can get stuck at the level of a model score while overlooking the decision the system supports or the people who rely on its output. This guidance helps organisations bring AI-related risks into their wider risk management work, from making decisions about potential uses to monitoring deployed systems.

What you are buying

Start by describing a credible harm in a specific use case, including who could be affected and who can act on it. Check whether your existing escalation and review process can handle changes to the model, data or way the system is used.

Frequently asked questions

Can we use this guidance with an existing risk register?

Yes. You can use it to bring AI-related risks into your existing process. Keep each entry tied to a specific use, the people affected and the team responsible for reviewing it.

Is reviewing the model's accuracy enough for an AI risk assessment?

Accuracy is one input. Consider the consequences of an error in the actual workflow, who relies on the result and how any change in use will trigger a review.