General

The process is the systematic use of policies, procedures and practices for communication and consultation, setting the context, and for the assessment, treatment, monitoring, review, recording and reporting of risk, and belongs inside management and decision-making rather than beside it (ISO 31000:2018, 6.1). The step before 6.1 supplies its input, so run both together.

What it asks

The process is the systematic use of policies, procedures and practices for communication and consultation, setting the context, and for the assessment, treatment, monitoring, review, recording and reporting of risk, and belongs inside management and decision-making rather than beside it (ISO 31000:2018, 6.1). For AI it is run for each AI system and tied to that system's life cycle, at strategic, operational, programme and project level, and allows for the way AI systems change over time.

Where generative AI changes the answer

A generative AI system can change behaviour between two risk reviews, so the process has to be triggered by a model or corpus change as well as by the calendar.

Reading it in practice

The process for generative AI needs two triggers, not one: the calendar review and a change event. A vendor model update, a new retrieval corpus, a new tool integration or a new class of user is a change event. Write the triggers into the process before the draft publishes, because after publication the first audit question will be how the organisation knew a system had changed.

Who it binds

The AI risk owner, the assessment team and the business owner of each generative AI use. For 6.1 the named role is the owner of documented AI risk management process applied per AI system.

Evidence an auditor normally samples

• Documented AI risk management process applied per AI system

The test you can run

Take documented AI risk management process applied per AI system and trace it to the decision it supported; if the trail stops before a named person, the control is not evidenced.

Common failures

• Risk assessed once at project approval and never during operation

Before the draft publishes

For 6.1, run the step once, by hand, on one live generative AI use, and keep the working paper. A method that has never been run on a real use is a template. The first artefact to produce is documented AI risk management process applied per AI system.

Where this sits against the draft

6.1 carries the risk management process part of the draft's published scope, which names risk sources, the stakeholders exposed to them, risk analysis, risk treatment and controls. When the draft publishes, compare its wording on 6.1 (general) with this record and note the change in the register.