The European Parliament has adopted its negotiation stance on the AI Act with an overwhelming majority, paving the way for the world’s first rules on Artificial Intelligence. These regulations will shape the final legislation and ensure that AI developed and utilized in Europe adheres to EU rights and values, encompassing human oversight, safety, privacy, transparency, non-discrimination, and social and environmental well-being.

Despite numerous proposed changes, the regulatory architecture identified by the Commission remains valid, aiming for a balance between innovation and the protection of fundamental rights.

The regulations employ a risk-based approach, prohibiting AI systems that pose an unacceptable risk to people’s safety. Intrusive and discriminatory uses of AI are also expanded to the list of banned practices. The use of live facial recognition is prohibited, with exemptions considered for national security and law enforcement. High-risk applications of AI, such as those influencing voter influence and large-scale recommender systems, will be subject to specific regulations. Generative AI systems like ChatGPT must comply with transparency requirements, register in the EU database, and mitigate potential risks. The legislation includes exemptions for research activities, promotes open-source AI components, and encourages the establishment of regulatory sandboxes.

The negotiations for the final text will proceed with the Council of the European Union, and the AI Act’s publication is anticipated by the end of 2023, to be enforced within two years.