A human reviewer approves thousands of AI recommendations without time to inspect evidence. What is the main governance concern?

1. Oversight quality is irrelevant once an AI system has launched.
2. The problem can always be solved by renaming the button reviewed.
3. Any human click guarantees every recommendation is correct.
4. The review may be a nominal gate rather than meaningful oversight.