Explainability

• Provide methods of explainability of black-box classification products, understandable to a non-expert audience
• Provide third-party audit of the AI solution

Ethical considerations relating to data limitations, fairness and bias

• Help identify data limitations
• Evidence that the AI system fits the requirements of the Data Ethics Framework
• Evidence an approach to eliminate bias, ethical issues, or other safety risks
• Provide software to allow the solution to be checked to detect bias
• Evidence audit and assurance with respect to intrinsic attributes such as considerations of fairness and mitigation of bias

Concept drift

• Provide software to help ensure solution does not drift from its intended outcome

Documentation

• Evidence of assurance processes

Lifecycle management

• Third party solutions to assess how accuracy and integrity is sustained over time
• Testing services for monitoring the efficacy of the model once deployed