Ethical maintenance
It is crucial to monitor that the AI system that fulfilled the Trustworthy AI requirement at launch, continues to do so over time. We call this “Ethical Maintenance”.
Guidelines
Use existing or define frameworks considering operational, financial, and reputational risk;
Guidelines
Perform the risk assessment of AI and identify the “opportunity cost of harm” for the organization (cost, revenue loss);
Guidelines
Determine the appetite of the organization for that risk in terms of money and conclude an appropriate budget level from a risk/benefit view;
Guidelines
Based on that assessment use this budget for an inspection of the AI asset in an RCM approach;
Guidelines
After inspection, feedback results, and potential impact and the likelihood of AI failures, e.g., false positives on the economic result;
Guidelines
Assess whether the current budget and depending factors are sufficient to manage the potential “opportunity cost of harm” and derive if further benefits/mitigations can be realized by expanding the inspection budget.