The European Parliament’s civil liberties (LIBE) and internal market (IMCO) committees have strongly supported draft legislation that establishes a risk-based framework for regulating artificial intelligence (AI) applications. In a vote held this morning, the committees approved the compromise reached with EU Member States by 71-8, with seven members abstaining. This agreement was the result of extensive three-way ‘trialogue’ discussions that took place at the end of last year.
The proposed EU AI Act, initially introduced by the European Commission in April 2021, aims to set comprehensive rules for AI developers based on the capabilities of their models and the intended application of AI. The Act outlines prohibited uses of AI—such as social scoring—and delineates regulations for high-risk applications in sectors like education, healthcare, and employment. These regulations include mandates regarding data quality, testing, and risk assessment. Additionally, the legislation implements transparency measures for general-purpose AI systems and applications, including tools like deepfakes and AI chatbots.
Most AI applications will be classified as ‘low risk,’ thereby excluding them from the law’s ambit. The framework also proposes the creation of regulatory sandboxes at the national level, allowing developers to design, train, and test high-risk applications within a supervised “real-world” context.
The initial proposal for the AI rulebook did not generate much excitement when unveiled three years ago. However, the emergence of generative AI over the past year has shifted focus onto the plan, deepening divisions among the bloc’s lawmakers. Members of the European Parliament (MEPs) have pushed to ensure the regulations encompass powerful general-purpose AIs, while some Member States—particularly France—advocate for exemptions for advanced AIs to nurture national digital champions.
The exhaustive trialogue negotiations in December resulted in a compromise text that still incorporates some provisions pertinent to general-purpose AIs, which has sparked ongoing opposition from certain governments. Recently, it appeared that these rifts could jeopardize the bill's progress. Nevertheless, a crucial vote by Member States on the compromise text earlier this month suggests that the EU is poised to finalize its landmark AI regulation in the coming months.
Despite this progress, the draft law must still navigate several final steps before adoption: a plenary vote in Parliament is scheduled for the coming weeks, where MEPs will formally consider the legislation, followed by a final endorsement from the Council.
While these concluding steps are expected to proceed smoothly, any disruption at this stage could significantly hinder the bill, particularly as parliamentary elections approach and the current legislative period comes to an end—creating a time-sensitive environment for decision-making.
The strong support from both parliamentary committees, which have rigorously scrutinized the legislative proposal over the years, signals that MEPs will likely maintain the substantial majority needed for passage. This could pave the way for the law to be enacted later this year, with the initial provisions—banning prohibited practices—expected to take effect six months afterward.
The EU plans a phased implementation of the Act, with legal requirements gradually ramping up for relevant developers between 2024 and 2027. Notably, the EU allows nine months for a code of practice to be established following the law’s enactment and 12 months for regulations specific to general-purpose AI, including governance rules, to take effect. The law will be fully applicable 24 months after it comes into force, although obligations for high-risk systems will follow a longer implementation timeline of up to 36 months.
Despite the anticipated approval of the parliamentary plenary vote, some dissent remains. For instance, the Pirate Party has declared its opposition, criticizing what its MEPs describe as a “flawed” law. Pirate Party MEP Marcel Kolaja, a member of the IMCO committee, expressed concern that national governments have diluted the AI Act's efficacy, leading the party to withhold support.
Another Pirate Party MEP, Patrick Breyer, a LIBE committee member, highlighted the potential risks associated with the Act, stating, “The EU’s AI Act opens the door to permanent real-time facial surveillance.” He cautioned that the law could permit public spaces across Europe to be subjected to biometric mass surveillance based on existing European arrest warrants, warning that this could cultivate a culture of mistrust and usher in what he termed a “dystopian future” characterized by pervasive surveillance.
Ranking
