Risk-Based Regulation

Tiering AI oversight by level of risk.

What characterizes a risk-based approach to AI regulation, as seen in frameworks like the EU AI Act?