What is a requirement for AI systems categorized as 'High Risk'?

For AI systems that are categorized as 'High Risk,’ conformity assessments are a crucial requirement. This entails a systematic evaluation to ensure that the AI system meets specific regulatory standards and requirements designed to mitigate risks associated with its functionality. High-risk AI applications can have significant impacts on individuals and society, which is why thorough assessments are mandated to confirm that systems operate safely, ethically, and in compliance with governance frameworks.

These assessments typically involve checking adherence to safety standards, regulatory requirements, and the ethical use of data. This rigorous process is intended to ensure that particularly sensitive applications — such as those involved in critical infrastructure, law enforcement, or healthcare — are reliable and do not pose undue risks to users or broader society.

The other choices reflect aspects that do not align with the specified requirements for high-risk AI systems. For instance, the absence of documentation is contrary to regulatory compliance principles. Third-party assessments can be beneficial but are not the sole requirement imposed on high-risk AI systems, and suggesting that deployers have fewer obligations directly undermines the need for comprehensive oversight associated with high-risk categorizations. Thus, the emphasis on conformity assessments is central to governance frameworks for high-risk AI.