EU Artificial Intelligence Act

The European Union’s draft Artificial Intelligence (AI) Act, reached a political agreement in December 2023, represents a significant milestone in global AI regulation and had been finalized in March 2024. Introduced by the European Commission in April 2021, it is the first binding worldwide horizontal regulation on AI, setting a common framework for the use and supply of AI systems within the EU. The Act introduces a risk-based approach to AI regulation, categorizing AI systems into different levels of risk and tailoring requirements and obligations accordingly.

Key provisions of the AI Act include:

• Prohibition of Certain AI Practices: Some AI practices are deemed unacceptable due to their significant risks to safety, livelihoods, and rights. These include manipulative subliminal techniques, systems exploiting vulnerabilities of specific groups, social scoring by public authorities, and real-time remote biometric identification in publicly accessible spaces for law enforcement, except under specific conditions.
• High-Risk AI Systems: AI systems that could have a detrimental impact on people’s health, safety, or fundamental rights are subject to stringent requirements before they can enter the EU market. This includes systems used in critical areas like law enforcement, education, and employment, among others.
• Transparency for Limited-Risk AI Systems: AI systems that pose limited risks, such as chatbots, must meet transparency obligations to inform users of their interactions with an AI system.
• Minimal Regulations for Low-Risk AI: AI systems posing minimal or low risk are not subject to additional obligations beyond existing legislation like the GDPR.

The AI Act outlines specific rules for general-purpose AI (GPAI) models, introducing more stringent requirements for models with “high-impact capabilities” that could pose systemic risks. It establishes a governance structure, including a European Artificial Intelligence Board and national supervisory authorities, to enforce the Act. Violations can lead to substantial fines, up to €30 million or 6% of the total worldwide annual turnover, depending on the severity of the infringement.

The legislative process has seen broad consultation and amendment, including input from the European Economic and Social Committee, national parliaments, and various stakeholders. Despite the challenges of defining AI systems, addressing their risk levels, and ensuring effective enforcement, the Act aims to balance innovation with consumer protection, fundamental rights, and safety.

As the AI Act approaches its final stages of approval, it underscores the EU’s commitment to a human-centric approach to AI, promoting the technology’s benefits while addressing the risks associated with its use. This legislative framework could serve as a model for AI regulation worldwide, reflecting the EU’s leading role in digital governance.