When is human oversight of an AI program typically mandated?

Human oversight of an AI program is typically mandated when the program has an impact on individual rights. This is rooted in ethical considerations and regulatory frameworks aimed at protecting individuals from potential harm that could arise from AI decisions. When AI systems are involved in processes that may affect personal rights—such as in legal, medical, or employment contexts—it's crucial to have human intervention to ensure accountability, transparency, and fairness.

The rationale for this oversight includes the need to mitigate risks associated with bias, discrimination, and erosion of privacy, among others. For example, if an AI system is used to determine loan eligibility or monitor employee performance, human oversight helps safeguard against decisions that could adversely affect people's lives without adequate justification or recourse.

In contrast, the other options do not align as strongly with the principle of mandated human oversight. An AI system being overly efficient, performing flawlessly, or being financially feasible does not inherently necessitate human involvement. Efficiency or flawless execution may not present the same ethical implications as a system that could potentially violate someone's rights. Therefore, human oversight is fundamentally tied to the implications of decisions made by AI systems on individual liberties.