The European Union reaches an agreement in principle to regulate artificial intelligence systems like ChatGPT

taketonews

After more than 18 hours of negotiation, the European Union (EU) still has not reached a definitive agreement on the new law that will regulate the artificial intelligence (AI). However, the institutions meeting in Brussels have closed a preliminary agreement on the rules that will be applied to the founding models, the technology that supports systems such as ChatGPT.

This part of the regulation It was one of their main conflict points. A couple of months ago there was a consensus on the need to apply a series of strict criteria on the foundational models, but France, Germany and Italy They changed their position and pushed to water down those regulations and leave self-regulation in the hands of large companies such as Microsoft, OpenAI either Google.

During the first six hours of the meeting, the delegates of the European Comissionof the European Parliament and the member states – represented by the Spanish presidency of the Council of the EU— reached a preliminary agreement of which, for now, no further details are known. Dozens of civil society organizations have warned that not establishing clear obligations and opting in favor of the giants of the sector will entail no small risks.

Pressure and a lot to close

The final three-way negotiations – known as trilogues – began on Wednesday at 3pm and were given until 8am this Thursday, when a press conference was to begin to give more details of what was agreed. Those schedules were, to say the least, optimistic. After more than 18 hours, there are still “many things” to close, according to negotiation sources told EFE. Thus, of the 22 key points of the law, there are still a dozen to be agreed upon.

The EU is negotiating against the clock, since the will of the Spanish presidency was to close this pioneering law before the end of the year. However, the pressure to reach an agreement on so many crucial aspects after long hours of technical negotiations and a sleepless night is causing discomfort with the political leadership, according to internal sources telling EL PERIÓDICO. Civil society organizations also explain to this newspaper that they see it as dangerous for the most important decisions regarding fundamental rights to be taken under a situation of fatigue and ask to move the meeting to another day.

facial recognition

Among the aspects still to be closed, some especially conflictive ones stand out, such as the possible prohibition of facial recognitionbut also the criticized exemption that would allow the police use the biometric surveillance in certain contexts.

European governments want security forces to be able to use these surveillance methods in real time and in public spaces – with prior judicial authorization – to, they say, prevent crimes such as terrorismhe sexual abuse or for the protection of critical public infrastructure.

Related news

The final agreement could opt to allow this exception, which worries civil society. “There is no level of mass surveillance that is acceptable, as it goes against the Fundamental rights“, he explained to EL PERIÓDICO Ella Jakubowskapolitical advisor of the organization European Digital Rights (EDRi).

Another point still to be debated is the article that aims to allow the 27 Member States to export AI systems of “unacceptable risk” or “high risk” to third countries that are not part of the EU. These two categories will be prohibited or very limited in the EU, as legislators understand that their use can represent a danger for the people.



ttn-24