General

Risk treatment selects and implements options for addressing risk in an iterative cycle of formulating and selecting options, planning and implementing treatment, assessing its effectiveness, deciding whether remaining risk is acceptable and treating further if not (ISO 31000:2018, 6.5.1); for AI the cycle should continue through operation because treatments applied to a learning system can lose effect. The step before 6.5.1 supplies its input, so run both together.

What it asks

Risk treatment selects and implements options for addressing risk in an iterative cycle of formulating and selecting options, planning and implementing treatment, assessing its effectiveness, deciding whether remaining risk is acceptable and treating further if not (ISO 31000:2018, 6.5.1); for AI the cycle should continue through operation because treatments applied to a learning system can lose effect.

Where generative AI changes the answer

Treatment for generative AI often starts with the cheapest option, which is to narrow the use, and that option is usually available and rarely taken.

Reading it in practice

The general step for treatment asks for a decision, not a list. For each generative AI risk the record should say whether the organisation will avoid the use, change it, transfer part of it by contract or accept it, and who decided. Acceptance is a legitimate answer when it is written, reasoned and owned. It is not legitimate when it is the absence of a decision.

Who it binds

The AI risk owner, the assessment team and the business owner of each generative AI use. For 6.5.1 the named role is the owner of treatment cycle records per AI system including effectiveness checks.

Evidence an auditor normally samples

• Treatment cycle records per AI system including effectiveness checks

The test you can run

Compare treatment cycle records per AI system including effectiveness checks with the use as it runs today; the gap between the two is the finding to record.

Common failures

• Treatment applied once with no check that it still works

Before the draft publishes

For 6.5.1, 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 treatment cycle records per AI system including effectiveness checks.

Where this sits against the draft

6.5.1 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 6.5.1 (general) with this record and note the change in the register.