In AI governance, which approach provides human oversight to mitigate bias and errors in critical decisions?

• A. Data quality in AI security
• B. TEVV in AI system development
• C. Human-in-the-loop in AI
• D. Chairperson

Answer: (C)

Human-in-the-loop provides a direct mechanism for human oversight in AI-driven decisions. In high-stakes scenarios, letting a person review, approve, or override automated outputs adds judgement, domain context, and ethical consideration that models alone often lack. This oversight helps catch biased data patterns, unintended consequences, and edge cases the system might misinterpret, preserving accountability and trust in the decision process.

Data quality in AI security is essential for reliable inputs, but it doesn’t by itself ensure ongoing human monitoring of decisions. TEVV focuses on verifying that the system meets requirements and behaves as intended, which is about validation and testing rather than real-time oversight. A chairperson is an organizational leadership role; while important for governance, it does not function as a built-in mechanism to mitigate bias or errors in critical AI decisions.