EU Artificial Intelligence Act: Unveiling the Framework of the Leaked Consolidated Texts
Introduction
On January 22, 2024, the public witnessed a significant milestone in the shaping of the EU Artificial Intelligence (AI) Act as two distinct versions of a consolidated text surfaced online. These leaked documents provide a glimpse into the ongoing negotiations, offering invaluable insights into the potential contours of the final AI Act.
Unveiling the Consolidated Texts
Euractiv Technology Editor Luca Bertuzzi released a comprehensive four-column document, spanning 892 pages, juxtaposing the original European Commission proposal with the mandates of the European Parliament and the Council. The fourth column serves as a draft consolidated agreement, highlighted with redlines, illuminating areas of convergence and divergence.
Shortly thereafter, European Parliament Senior Advisor Laura Caroli shared an alternative perspective, offering a consolidated 258-page document. These two documents collectively provide a comprehensive understanding of the proposed AI Act.
Navigating the Tight Timeline
The release of the consolidated texts comes amidst a pressing timeline for finalizing the AI Act. With parliamentary elections scheduled for June 2024, EU member states are tasked with analyzing and agreeing upon the extensive text within a limited timeframe. National delegates are likely to prioritize key articles, given the time constraints.
Key Dates and Implementation Phases
Upon its adoption, the AI Act would enter into force 20 days after publication in the EU Official Journal. Its various provisions would be implemented in a phased manner:
– Prohibitions on unacceptable risk AI systems: 6 months after entry into force
– Obligations for high-risk AI systems: 36 months after entry into force
– Obligations for providers of general purpose AI models, provisions on penalties for general purpose AI systems, and appointment of member state competent authorities: 1 year after entry into force
– European Commission implementing act on post-market monitoring and list of elements for monitoring: 18 months after entry into force
– Classification rules and procedures, member state rules on penalties, and establishment of AI regulatory sandboxes: 24 months after entry into force
Governance, Knowledge, and Training Highlights
The leaked texts emphasize the significance of AI governance, appropriate training, qualifications, and skills. Several recitals and articles address these aspects:
– Recital 9(b): Focus on “AI literacy” to ensure compliance and enforcement
– Recital 48: High-risk AI systems should be designed for human oversight and impact mitigation throughout the lifecycle. Oversight personnel must possess the necessary competence, training, and authority.
– Recital 58: Deployers of high-risk systems must monitor, provide instructions for use, and maintain records of such actions. Appropriate training is required for those tasked with these obligations.
– Recital 60(q): Providers of general purpose AI systems should continuously assess and mitigate systemic risks, including implementing risk-management policies with accountability processes.
– Articles 3, 4, 9, 14, 29, 33, 54, 55, and 59: Various articles touch on AI governance and knowledge obligations, including the provision of adequate resources to national competent authorities.
Continued Developments and IAPP Involvement
The IAPP will remain vigilant in monitoring the progress of the EU AI Act, providing timely updates and in-depth resources to assist organizations in understanding the requirements and obligations of the legislation. As the negotiations continue, the IAPP will be at the forefront, disseminating the latest information and offering expert insights on this transformative piece of legislation.
Conclusion
The leaked consolidated texts of the EU AI Act offer a profound glimpse into the ongoing negotiations and the potential shape of the final legislation. With its emphasis on AI governance, appropriate training, and phased implementation, the AI Act aims to strike a balance between fostering innovation and mitigating risks associated with AI technologies. The IAPP stands ready to provide ongoing support and guidance as the EU AI Act takes shape, ensuring that organizations are well-equipped to navigate the regulatory landscape and harness the transformative potential of AI.