The European Parliament approves the negotiation of the world’s first law on artificial intelligence | technology

On Wednesday, the European Parliament approved by a large majority its final position on artificial intelligence (AI) law, thus opening the door to discussions that will begin immediately with EU countries and the Commission, the so-called experimental models. The aim is to complete the negotiations during the Spanish presidency of the European Union, which has already declared this regulation one of its priorities. All this so that by 2026 at the latest, but perhaps even sooner, Europe will become the first region in the world to have standards in place regulating a rapidly evolving technology that promises as much as it frightens because of its infinite potential to change the world. the society.

The Internal Market Commissioner, Thierry Breton, immediately held a comfortable vote – 499 in favour, 28 against and 93 abstentions – making the European Parliament, as he stated, “the first parliament in the world to vote on a comprehensive regulation for artificial intelligence”.

“Artificial intelligence is already an important part of our lives and it has raised a series of doubts in terms of ethics, auditing, innovation and the need to find the right regulatory framework,” said the President of the European Parliament, Roberta Metsola, celebrating the progress. By voting on the parliamentary position that “Europe takes the initiative and does it in our own way and responsibly.”

“If there is something we cannot accept compromises on, it is in the fact that every time technology advances, it must go hand in hand with our fundamental rights and democratic values,” Metsola stressed.

MEPs’ final position is tougher than the one already agreed by the states in December, and which will now have to be negotiated for a single final text: plus so-called social credit systems, which had already been rejected in the Commission’s original version. In the proposal, the European Parliament also wants to ban “intrusive and discriminatory uses of artificial intelligence”, especially remote biometric identification systems, and profiling based on sensitive characteristics such as gender, race, ethnicity, religion or political orientation. It also provides for the veto of predictive policing to assess the risk of a person or group of persons committing an offense or offense (based on the profile, location or criminal past of said persons), as well as emotion recognition systems, for example through police or agents borders or in workplaces and schools.

“The world has matured so that this use-and-risk approach makes more sense for the future.”

Margaret Vestager

The European People’s Party (EPP) at the last minute — thus breaching, according to the law’s negotiators, an agreement not to add anything not discussed — introduced an amendment to restore the exceptions originally provided for and allow law enforcement to use biometric recognition systems, such as facial recognition, in Public locations in real time for very specific situations, such as the search for a missing child or a terrorist attack. It is a similar position to the one approved by the Council in December. Finally, it was rejected in plenary, despite the fact that during the debate on the law, on Tuesday, popular MP Jeroen Lenners defended the law as a sign that, when it comes to artificial intelligence, “you don’t have to just talk about how to protect citizens from intelligence.” artificial intelligence, but also how AI can protect citizens.” The final proposal continues to allow the use of “previous job“From this kind of technology to those few cases, as long as there is a court order for that. A sufficient procedure, according to the rapporteurs of the parliamentary motion, the Italian Brando Benifi (S&D) and the Romanian Dragos Tudorache (Renewal) to handle these cases.

‘Unacceptable’ risks

Generative AI systems—capable of creating original audio, text, or image content based on observation of other data—such as ChatGPT, were little known when the first legislative steps were taken, but are now on the minds of everyone, including their creators, who are strong contenders for a clear delineation. to use it. MEPs have added extra layers of security and transparency for the developers of these foundational models in this regulation that is based not on regulating specific technologies, but on how they are used: the law creates categories ranging from “unacceptable” risks (prohibited by legislation) to other acceptable minors, though From being subject to strict controls so that they do not affect the freedoms and rights of citizens, especially the second category in terms of “high risks”, where generative artificial intelligence enters.

The applicants of these constituent forms must, according to the proposal of the MEPs, ensure “strong protection” of citizens’ fundamental rights, for which they must “assess and mitigate risks, comply with European design, information and environmental requirements” and be “registered” in the EU database. In addition, generative models such as ChatGPT must meet additional “transparency” requirements: thus, they would have to make it clear that the content was generated by an AI system, which would require a watermark to distinguish it. Likewise, they must design their models in a way that prevents them from creating illegal content, and in order to protect copyrights, they must publish the data used to train these systems, in a way that if the author believes that your rights have been violated through the use of algorithms for your material, you can resort to To the existing legal channels in the European Union to claim or claim compensation.

We want AI content to be known and deep fake so it doesn’t poison our democracy.

Brando Benefi

We want AI content to be discoverable to users Deep fakes Do not poison our democracy,” Benifee noted. During the debate on Tuesday, the European Commission Vice-President for Digital and Competition, Margrethe Vestager, highlighted the importance of a regulatory framework for AI because, as she said in plenary session of the European Parliament, it allows “creating trust and legal certainty. Thus, he added, “Great things can be done with AI if you can trust it.”

The Danish company, which at the end of May launched with the United States a proposal to create a voluntary “code of conduct” for the generative AI industry under the G7 that serves as a “bridge” until European law and others in other parts of the world are up and running, he says, now it is The “right time” to legalize this technology the way the European Union deals with it. “The world has matured so that this use-based, risk-based approach makes more sense going forward,” he said. For his part, Commissioner Breton considered that Europe is the continent “most capable of implementing this regulation in a balanced, transparent and proportionate manner.”

“We have the possibility of gradually setting a global standard,” added the Frenchman, who on Wednesday announced his intention to celebrate his first triathlon tonight in Strasbourg because “there is no time to lose. I count on the next Spanish EU presidency to work on this great issue to come up with a final text.” At the end of the year,” he said. Benife even declared himself “sure” that this deadline would be possible even if there was a change of government in Spain on July 23, because, as the Italian Social Democrat asserts, the Spanish presidency showed itself “very committed” to this matter.

However, community sources acknowledge that it will not be easy to reconcile a vision that places more emphasis on the national security of states with one on protecting citizens’ basic rights in high-risk areas. But they also declare that they are optimistic given the context in which the negotiations will take place: growing concern about the risks that generative AI may entail, expressed even by its developers, many of whom have explicitly called for regulation, and messages that other regions and countries in the world are already willing to put in place safeguards technology whose potential is hard to imagine. The main idea, they noted, was that neither Europe nor anyone else could not organize such an important issue. In fact, Benifi argues that there is “enough appetite” even for the times and that the AI ​​law comes into force, in whole or in part, earlier than expected by the commission, which expects it to be fully implemented in 2026.

You can follow country technology in Facebook And Twitter Or sign up here to receive The weekly newsletter.