The European Union (EU) is poised to finalize a groundbreaking agreement on rules governing artificial intelligence (AI), capping months of intricate negotiations, particularly concerning the oversight of generative AI models like ChatGPT.
The urgency to address AI regulation heightened following the notable emergence of ChatGPT last year, showcasing the rapid advancements in AI capabilities as ChatGPT has underscored the transformative potential of AI across various domains, including work and healthcare.
Proponents argue that AI stands to benefit humanity significantly. However, concerns about the societal risks it poses have intensified, with apprehensions of potential unprecedented chaos.
European governments are determined to rein in major tech players by wielding a robust legal framework aimed at safeguarding the rights of EU citizens, particularly in areas of privacy and data protection. The European Commission initially proposed an AI law in 2021, outlining a risk-based approach to regulation, where the level of obligation correlates with the potential risk to citizens' rights or health.
Negotiations on the final legal text commenced in June, with recent weeks witnessing a heated debate on how to regulate general-purpose AI, such as ChatGPT and Google's Bard chatbot, posing a last-minute threat to the talks. Some member states expressed concerns that excessive regulation could stifle innovation and hinder the emergence of European AI giants to rival those in the United States.
Negotiators from the European Parliament and EU member states are set to convene on Wednesday, with discussions anticipated to extend into the evening. Although there is no strict deadline, senior EU figures emphasize the need to finalize the law by the end of 2023.
Despite the progress made, potential stumbling blocks still exists, and there are warnings that an agreement may not be reached. Even if a political agreement materializes, additional meetings may be required to finalize the law's technical details. If negotiators reach a consensus, the law is not expected to come into force until 2026 at the earliest.
A key point of contention revolves around the regulation of foundation models—versatile models designed to perform various tasks. France, Germany, and Italy advocate for their exclusion from regulation, favoring adherence to codes of conduct.
Another contentious issue is remote biometric surveillance, specifically facial identification through public camera data. The EU Parliament calls for a complete ban on "real-time" remote biometric identification systems, while member states oppose such a ban. Concessions may be considered in this area in exchange for agreements in other domains.
Despite these challenges, the EU's internal market commissioner, Thierry Breton, expressed optimism about reaching an agreement. Acknowledging significant progress on both sides, Breton pointed to the mandate given to Spain, currently holding the rotating EU presidency, to conclude negotiations on Wednesday.