AI systems may be subject to performance deterioration over the medium and long term. In addition, the business, operational, IT, and regulatory environments and stakeholder pressures will change over time. These processes may jeopardize system performance or lead to unacceptable risks.
The AI System Owner should ensure that the organization conducts regular comprehensive reviews of the AI system (health checks) to ensure that the AI system aligns with the organization’s values and risk tolerance.
The AI System Owner should ensure that the organization defines, documents, and entrenches workflows and technical interfaces for reviewing
1) AI system use case,
2) AI system users,
3) AI system operational environment,
4) AI system technical environment,
5) AI system metrics,
6) AI system version control practices,
7) the AI system monitoring practices and
8) the AI system health check practices.