Comments on the Final Trilogue Version of the AI Act

Abstract

This paper provides a comprehensive analysis of the recent EU AI Act, the regulatory framework surrounding Artificial Intelligence (AI), focusing on foundation models, open-source exemptions, remote biometric identification (RBI), copyright, high-risk classification, innovation, and the implications for fundamental rights and employment.

The comments critically examine the regulatory approach to general-purpose AI systems, highlighting the new challenges and critiques regarding the regulation of foundation models, the functioning and limitations of open-source exemptions, and the concerns raised by RBI, especially in terms of privacy, function creep, and enforcement issues. It further delves into the copyright regime affecting AI, the criteria for high-risk classification, and the intersection with innovation, discussing how AI regulation aligns with existing sectoral regulations and impacts the AI value chain. The paper also addresses regulated self-regulation, safe harbors, the fundamental rights impact assessment, the right to an explanation, compliance timelines, and the preparation businesses must undertake to comply with new regulations.

By analyzing the regulatory framework's strengths and weaknesses, the paper argues for a balanced approach that considers the importance of minimum rules for foundation models and RBI, the potential for regulated self-regulation, and the need for international collaboration and standards. It emphasizes the importance of addressing compliance costs, especially for SMEs, and proposes strategies for improving alignment with sectoral regulations, enhancing European supervision of RBI, and fostering innovation while safeguarding fundamental rights and democratic values.

Keywords: AI, artificial intelligence, machine learning, AI Act, RBI, innovation, data protection