The European Parliament’s civil liberties (LIBE) and internal market (IMCO) committees have given strong backing to draft legislation aimed at establishing a risk-based framework for regulating artificial intelligence (AI) applications.
In an emphatic vote, the committees endorsed the compromise reached with EU Member States late last year, marking a significant step toward the adoption of the EU AI Act.
Originally proposed by the Commission in April 2021, the EU AI Act sets rules for AI developers based on their models’ power and intended purposes. It outlines prohibited uses of AI, including social scoring, and establishes rules for high-risk applications in areas like education, health, and employment. Transparency requirements are also imposed on general-purpose AIs, deepfakes, and AI chatbots.
The majority of AI applications are categorized as ‘low risk,’ falling outside the law’s scope. The legislation also introduces regulatory sandboxes at the national level, allowing developers to test risky applications in a supervised real-world environment.
While the Commission’s initial proposal did not garner much attention three years ago, the rise of generative AI has brought the plan into the global spotlight, sparking debates among EU lawmakers. Amendments were proposed to ensure the legislation covers powerful general-purpose AIs, while some Member States, led by France, sought regulatory exemptions for advanced AIs to support national champions.
After lengthy trilogue talks in December, a compromise text was reached, but divisions persisted. The recent Member State vote on the compromise text cleared a major hurdle, making the adoption of the flagship AI rulebook likely in the coming months.
Despite some remaining opposition, especially from the Pirate Party, which considers the law “flawed,” today’s strong endorsement by the two parliamentary committees indicates a probable majority support during the upcoming plenary vote. This positive momentum suggests the law could be adopted and enter into force later this year, with the first provisions likely to apply in the second half of 2023.
The phased rollout of the EU AI Act is expected to see legal requirements increasing for developers between 2024 and 2027. Despite potential challenges, the legislation is poised to bring comprehensive AI regulation to the European Union.