Which term describes AI systems that have the potential to harm safety or rights of individuals?

The term that best describes AI systems that have the potential to harm the safety or rights of individuals is "High risk." This classification is typically used to denote AI applications that pose significant threats to public safety, personal safety, fundamental rights, or welfare. These systems are often subjected to stricter regulatory oversight and compliance requirements due to the potential for serious negative impacts if they fail or are misused.

For example, systems involved in healthcare decisions, law enforcement, or critical infrastructure management are often considered high risk, as any malfunctions or biases in these applications can lead to severe consequences for individuals and society as a whole. Understanding this classification is crucial for governance, as it informs the need for robust evaluation frameworks, monitoring, and accountability measures to mitigate potential harms associated with these AI systems.

In contrast, other risk categories such as "Limited risk," "Minimal risk," or "Unacceptable risk" reflect different levels of concern regarding the implications of AI applications, but they do not convey the same urgency and seriousness associated with high-risk systems.