The lack of agreement in the European Union endangers the regulation of artificial intelligence

Barcelona

11/14/2023 at 06:17

CET


Facial recognition is one of the most controversial points: “There is no level of mass surveillance that is acceptable, as it goes against fundamental rights,” says Ella Jakubowska, political advisor at the organization European Digital Rights (EDRi).

The European Union (EU) has been working for years on a pioneering regulation of artificial intelligence (AI). The project, which aims to prohibit the most dangerous uses of this technology, is in the last phase of the legislative process. A final agreement should be closed in the coming days. However, the tense lack of agreement between the parties threatens to put all the law at risk.

On June 14, the European Parliament gave the green light to first comprehensive AI legislation of the world. With an overwhelming majority of 499 votes in favor, the chamber agreed to demand the prohibition total of the mass surveillance with real-time biometric identification systems and establish a series of restrictions on the use of Generative AI like the well-known ChatGPT.

However, the text approved then is not the definitive one. The final wording of the law must be agreed upon with the European Comission and the European Council —made up of the political leaders of the 27 Member States—, a negotiation three-way that could weaken the project to the point of derailing it. This process, known as the trilogues, is evidencing a visible lack of harmony on how to regulate AI.

Biometric surveillance

One of the most controversial points is the systems of facial recognition. Although the European Parliament agreed to prohibit its use in real time, in recent days MEPs have abandoned this total restriction in exchange for other concessions. Thus, according to the portal Euractivthe use of this technology by the police and in specific cases such as locating a missing person or to prevent a terrorist attack.

This turn deeply worries civil society organizations. “There is no level of mass surveillance that is acceptable, as it goes against the Fundamental rights“, he explains to EL PERIÓDICO Ella Jakubowskapolitical advisor of the organization European Digital Rights (EDRi). The expert denounces that Parliament has made “enormous concessions” on a democratic agreement already voted on and warns of the “danger” that, if approved in this way, the European law could legitimize the biometric surveillance worldwide.

The AI will continue to be used to monitor people migrants in the borders Europeans, an application that arouses suspicion more outside than inside the negotiations. “It is sad that the prohibition of these systems to experiment on the most vulnerable is not even on the table,” laments Jakubowska. “Lack of humanity is built into AI law, it’s a missed opportunity.”

Dispute with ChatGPT

Another point of dispute is the so-called foundational models, the technological heart of generative tools such as bardof Googleeither ChatGPT and DALL-Eboth created by OpenAI and financed by Microsoft. A few weeks ago there seemed to be a consensus on the need to set stricter standards for the most powerful systems, an approach based on the potential risk which especially harmed those American business giants.

However, the approach displeases large EU countries such as France, Germany either Italywho have opposed any type of regulation of the foundational models. The governments of the three nations argue that these regulations will endanger the innovationbut its “self-destructive” rejection also responds to “pressure from the AI ​​industry,” according to Daniel Lauferpolitical analyst of the international organization Access Now.

The change in position so outraged MEPs that they walked out of last Thursday’s technical meeting in protest. It also outrages civil society. Laufer asks States to be more open to negotiating to force companies to be transparent. “We are talking about key rights for our democracies,” he warns in statements to EL PERIÓDICO.

Clock against

The initial forecast was that the political agreement on the AI ​​law would be closed on December 6. However, disagreements between the parties could postpone it. Spain He is pressing for an agreement to be reached before the end of the year and his presidency of the Council. If not, it will be the next presidency, managed by Belgium, that will have to tie up the loose ends before the European elections in June. Time is against a pioneering project that could end up being a dead letter.

ttn-25