What are the three approaches to human oversight of AI systems mentioned in the Ethics Guidelines for Trustworthy AI?

The three approaches to human oversight are human-in-the-loop, human-on-the-loop, and human-in-command, which are meant to support human decision-making while maintaining appropriate oversight of AI systems.

In the Ethics Guidelines for Trustworthy AI, these approaches describe different levels and modes of human involvement. Human-in-the-loop means humans are directly involved in the system's decision cycle. Human-on-the-loop means humans can supervise and intervene while the system operates, but are not part of every decision. Human-in-command means humans hold ultimate authority over the AI's overall goals and behavior, especially in terms of deciding when and how to use it.