A risk-based approach assigns regulatory requirements according to an AI system’s potential severity and likelihood of harm. Greater risks receive stricter controls. section 10.30
The approach distinguishes systems by the seriousness of their possible effects:
- Unacceptable-risk systems can be banned when they violate fundamental rights, including through social scoring or manipulation of vulnerable groups. section 10.30
- High-risk systems affect safety or fundamental rights, including AI used in hiring, education, law enforcement, and financial access. section 10.30
- High-risk systems require lifecycle assessment, risk management, documentation, record-keeping, human oversight, data governance, accuracy, nondiscrimination, and registration. section 10.30
- Lower-risk systems receive lighter treatment rather than automatically facing the controls imposed on high-risk systems. section 10.30
A prohibited children’s toy
A voice-activated toy that encourages violent behavior in children illustrates an unacceptable-risk application and could be banned. section 10.30
Figure 10.7 The four risk categories in the AI Act
Use the four-category overview to compare how regulatory attention changes with an AI system’s level of risk. section 10.30
