From Policy to Practice: Prototyping the EU AI Act's Human Oversight Requirements

Abstract

Article 14 AI Act mandates that high-risk AI systems are designed and developed in such a way that they enable effective human oversight. It establishes the requirement that human oversight be made possible by providers both through the design of the high-risk AI system as well as through organisational measures, identified by the provider and implemented by the deployer. This should ensure that the risks to health, safety and fundamental rights are effectively mitigated. 

This project intended to test the human oversight requirements for high-risk AI systems in the AI Act by gathering stakeholders to implement these requirements into prototype compliance documents. We collected feedback on these prototype compliance documents in order to determine best practices and policy recommendations. This report includes the compliance documents drafted by the participants of the policy prototyping workshop in its annex. In addition, the report also features reviewer feedback on those documents and the human oversight requirement contained in article 14 AI Act.

Keywords: AI Act, Human oversight, high-risk AI systems, policy prototyping, product safety, law enforcement, AI in education, AI in healthcare, Artificial Intelligence, AI governance, compliance, AI regulation