What could be the pivotal negotiations to agree on the world’s first legislation on the issue are on the agenda today. The EU Council is working with EU Parliament negotiators to reach a compromise on the future rule, but it needs to get over France, Germany, and Italy’s resistance to the foundation models.
Brussels – The EU Act on Artificial Intelligence, which could result in the first laws ever on the topic worldwide, is most significant today. However, there is no guarantee that the co-legislators of the EU Parliament and the Council will come to a consensus, as there is still a chance that the final round of trilogues, which is set for this afternoon (December 6), will not go through because of certain challenges that the rotating Spanish presidency of the EU Council is attempting to resolve. Speaking yesterday (Dec. 5) at the meeting of EU ministers in charge of telecommunications, Carme Artigas, the Spanish Secretary of State for Digitization and Artificial Intelligence and the rotating chair of the EU Council, said, “We have aligned our positions and we have enough flexibility to discuss with Parliament.”
Reiterating that “the important thing is to reach a political agreement,” Artigas made it clear that technical work would still need to be done in order to finalise the Regulations by the end of the legislature in the spring of the following year, even if a deal is reached between negotiators from its EU institutions today. Currently, there are three key chapters that still require extensive mediation: law enforcement, artificial intelligence generative models, and classification requirements for “high-risk” systems. The last of the three presents the biggest challenge because, for weeks, France, Germany, and Italy have opposed the Spanish presidency’s step-by-step plan on foundation models—generative artificial intelligence that has been trained on a wide range of generalised, label-free data—and its call for codes of conduct for the models’ development.
EU institutions are aware that, subject to the trialogues’ conclusion, certain sections of the Regulation will take effect immediately upon its adoption, while others—like the ones pertaining to high-risk artificial intelligence systems—would only become operative at the conclusion of a transitional period. For this reason, the Pact on AI was introduced on November 16 in an effort to encourage the industry to voluntarily commit to starting to implement its standards before of the legislative date, particularly with regard to generative Ia systems ahead of the European elections in June of the following year. To increase transparency and foster greater confidence, participating companies will sign commitment statements that are supported by specific ongoing or planned measures that the EU Commission will publicise. In advance of stakeholder discussions on draft concepts and best practices, which are anticipated to take place in the first half of 2024, a request for expressions of interest has been issued. The key organisations of the Covenant will be invited to publicly announce their initial pledges following the official approval of the AI Act.
The Fundamentals of the Artificial Intelligence Regulation in the EU
The European Commission first proposed a legal framework for artificial intelligence on April 21, 2021, to allow for the development and application of cutting-edge technologies within EU borders. It is the world’s first legislative attempt to create a regulatory framework for artificial intelligence systems. The executive body’s approach will be pursued by the Council and the European Parliament, and following modifications to align the positions of the two organisations, a risk ladder governing artificial intelligence applications on four tiers will be developed: Minimal (spam filters and AI-enabled video games), limited (chatbots), high (grading college and career examinations, organising resumes, determining the validity of testimony in court, and performing robot-assisted surgery) and unacceptable (anything that presents a “clear threat” to people’s rights, security, and way of life, such government “social scoring”).
There won’t be any involvement for the first level, and the last level will be completely prohibited. Artificial intelligence systems that use deliberate or covert manipulation, prey on people’s vulnerabilities, or are employed for social scoring will all be outlawed if they pose an intolerable risk to public safety. The amendment game is being played on the inclusion of remote biometric identification systems in publicly accessible spaces in this list, both in real time and later on, at the behest of MEPs. These systems use biometric categorization that creates databases of facial recognition using facial recognition software from social media or CCTV cameras, as well as sensitive characteristics like gender, ethnicity, citizenship, religion, and political orientation. Additionally, border management, businesses, and educational institutions are using emotion recognition software and predictive policing systems (based on profiles, location, or past criminal activity).
Lastly, the proposed EU AI law may include the establishment of a European Office for Artificial Intelligence, as EU Commission President Ursula von der Leyen noted during the global conference in the UK. The head of the EU executive clarified, “This Office could deal with the most advanced Ia models, with responsibility for supervision,” adding that it should adhere to the four principles she outlined in her speech on the global governance framework and enforce uniform regulations for the most advanced models in each of the 27 member states. Von der Leyen’s statement suggested that the EU Office for Artificial Intelligence “should also have a global vocation,” working with other organisations of a similar nature around the globe.
