The higher the risk to rights, safety, or society, the heavier the obligations.
The AI Act sorts systems by risk: unacceptable (banned), high (strict controls), transparency (limited duties), and minimal (free). The exam trap is confusing transparency obligations with high-risk rules.