Integrate management into the stages of the lifecycle
AI Ethics: Managing Risks and Impacts under ISO/IEC 42001:2023

The first step is to define:

• the purpose of the system (what decisions, what role for AI);

• stakeholders (users, affected individuals, business units, the Data Protection Officer, or the privacy officer);

• regulatory and standard-related requirements.

A preliminary risk and impact analysis makes it possible to classify the project by level of criticality, identify the main categories of risk, and determine whether a formal impact analysis is required. As the Amazon case demonstrates, this phase must include an assessment of the representativeness of the proposed training data and the potential biases it may contain....