For over two years, the European Union has been working on an ambitious project to regulate artificial intelligence, known as the AI Act. The main goal of this initiative is to promote innovation while ensuring the safety and rights of users. On Thursday, May 11, during a committee vote, Members of the European Parliament took a decisive step by approving a strengthened version of the AI regulation proposal compared to the initial draft from April 2021. This decision was made with a broad majority, reflecting the importance placed on this issue. Now, let's explore the five key points behind the AI Act.

The AI Act aims to establish a coherent legal framework for the commercialization and use of artificial intelligence. Its primary objective is to prevent potential abuses related to AI while maintaining an environment conducive to innovation. According to German MEP Svenja Hahn, quoted by Reuters, "We have managed to find a compromise that regulates AI in a proportionate way, protects civil rights, and boosts innovation and the economy." The adoption of such regulation would constitute a global first on this scale, marking a significant step in the field of AI regulation.

Europe proposes a regulation that integrates existing product safety regulations, while introducing control measures for artificial intelligence applications considered "high-risk." These "high-risk" applications span various fields such as critical infrastructure, education, human resources, law enforcement, and migration management. According to the text, companies developing these "high-risk" AI systems must meet several requirements, including ensuring human oversight, providing technical documentation, and implementing a risk management system. Each EU member state will be responsible for establishing its own regulatory body. No additional obligations will be imposed on applications that are not considered "high-risk."

During the vote, Members of the European Parliament approved the ban on the use of facial recognition in public spaces as well as predictive policing tools. These two examples illustrate the AI applications that Europe clearly wants to prevent from developing on its territory. Furthermore, the April 2021 text also targeted systems that establish social scores, biometric identification, subliminal manipulation, and targeting of vulnerable individuals. All of these uses of AI were covered by the proposed restrictive measures.

Regarding generative AIs, which have transformed the digital landscape and gained widespread popularity in recent months, new transparency measures will be imposed. Tools such as ChatGPT, Midjourney, or DALL-E will likely be required to disclose copyrighted content used to train their language models. Furthermore, these AIs should be subject to testing by independent experts in the future to assess potential risks related to security, health, fundamental rights, democracy, and the environment. These hard-to-avoid risks and potential abuses must be clearly specified in separate documentation. Additionally, compliance with the General Data Protection Regulation (GDPR) will be crucial in all cases.

After two years of negotiations, the draft of the regulation on the use of AI has been adopted by the European Parliament, marking an important step toward the adoption of a complete text. The next step in the legislative process is to confirm the result of this preliminary agreement during a plenary session of the commission in June. Subsequently, a new negotiation period with the Member States will follow, leading to the drafting and voting of the final text. Margrethe Vestager, Vice-President of the Commission, expressed her determination to waste no time and hopes to begin the first political negotiations before the summer, with the goal of concluding them this year. Are you a professional looking for a communication agency for your project? Schedule a meeting with our team to discuss your project and discover our tailored working method.