The Limited Times

Now you can see non-English news...

EU approves world's first artificial intelligence law

2023-12-09T22:47:23.459Z

Highlights: EU approves world's first artificial intelligence law. The States and the European Parliament overcome their differences on how to regulate the foundational models and red lines in terms of biometric surveillance systems. The final text will still have to be ratified by the two parties before entering into force, foreseeably at the end of 2026, although some parts will start working earlier. This text defines the obligations and rules that must govern a technology that is here to stay and that is completely transforming daily life, but that carries as many possibilities as risks, many of them not even imaginable yet.


The States and the European Parliament overcome their differences on how to regulate the foundational models and red lines in terms of biometric surveillance systems and close a provisional agreement that still needs to be ratified before it enters into force


The European Union will be the first region in the world to fully regulate the uses of artificial intelligence (AI). The States and the European Parliament have reached, almost at midnight from Friday to Saturday, and after three days of intense and tough negotiations, a provisional agreement, a final text that will still have to be ratified by the two parties before entering into force, foreseeably at the end of 2026, although some parts will start working earlier. This text defines the obligations and rules that must govern a technology that is here to stay and that is completely transforming daily life, but that carries as many possibilities as risks, many of them not even imaginable yet.

"The regulation aims to ensure that AI systems used in the EU are safe and respect fundamental rights and European values," the Spanish Presidency of the Council of the EU said on X (formerly Twitter). Closing this law, which aims to become a reference or standard for other regions beyond the EU borders, was one of the priorities set by Spain for its European semester, which ends this December.

The Twenty-Seven say they want to guarantee the possibilities and limit the risks as much as possible, for which they have negotiated a "future-proof" law that contains enough flexibility to be able to regulate functions or technologies that are currently unknown, or to adapt to the changes experienced by existing ones. But the devil, as always, is in the details, and the negotiations, "passionate", as they were described by witnesses of the endless tug-of-war that began on Wednesday and ended up close to midnight from Friday to Saturday, dragged on due to the standoff between states and MEPs – and sometimes between each other – over what is a risk or not and what exceptions and safeguards to put in place to ensure that individual fundamental rights are not violated. they take risks in order not to harm the economy or the interests of States. The fine print is yet to be known, but, after 36 hours of negotiations, one of the longest in living memory, all parties declared themselves satisfied with the "balance" achieved between the two objectives.

Learn more

The main artificial intelligence companies accept the European law, but demand that its application does not put a brake on it

"It is a very good law that will promote innovation and in a way that is compatible with fundamental rights," said the Secretary of State for Digitalisation and Artificial Intelligence, Carme Artigas.

Deal!#AIAct pic.twitter.com/UwNoqmEHt5

— Thierry Breton (@ThierryBreton) December 8, 2023

"The EU becomes the first continent to set clear rules for the use of AI," said Internal Market Commissioner Thierry Breton, one of the main promoters of the regulation and for whom the proposed law is "much more than a package of rules, it is a launchpad for European startups and researchers to lead the global race for AI".

Above all, Artigas said, because it gives "legal and technical certainty" to citizens and companies that will foreseeably save many legal actions. To do this, the regulation wants to have enough teeth so as not to remain a dead letter, for which it provides for a system of penalties, either a percentage of the total turnover of the offending company in the previous fiscal year or even an "even higher" predetermined amount. In addition, it establishes the creation of an independent supervisory body, an AI Office linked to the European Commission and which will be advised by a scientific panel and civil society.

The President of the Commission, Ursula von der Leyen, has welcomed a "pioneering regulation in the world". A "unique legal framework for the development of AI that can be trusted", the head of the European Executive greeted in X. "Europe has led and delivered," said the President of the European Parliament, Roberta Metsola, according to which the AI Act, as it is known in English, is a "forward-thinking and responsible legislation that imposes global standards".

Along with the satisfaction of closing an agreement that often seemed to be slipping away, there were also gestures of relief among the main people responsible for the negotiations, and not only because they could finally go home, but because, they all stressed, they were sure that they had done their homework well with many obstacles.

Among them, the thorniest were, as expected, the question of how to regulate general-purpose AI models (generative AI or foundational models) on which popular tools such as ChatGPT are based, as well as biometric surveillance systems (such as facial recognition), which in the end ended up being the most arduously negotiated point and prolonged discussions until this Friday. Due to the strong misgivings aroused by these technologies that potentially – and in some cases already really, as seen in some countries with little democratic muscle – allow state supervision and control that can collide directly with the fundamental rights of citizens.

Finally, the failure of the negotiations, which Spain insisted on concluding now, fearful of a timetable that, with European elections in six months, could jeopardise a law that has been drafted and discussed for a long time, has been avoided. The European Parliament welcomed the fact that the law "seeks to ensure that fundamental rights, democracy, the rule of law and environmental sustainability are protected from high-risk AI, while driving innovation and making Europe a leader in the sector".

According to MEPs, the final text includes its main red lines, confirming that several biometric surveillance systems that they considered unacceptable will be prohibited: biometric categorisation systems (by political, religious, philosophical beliefs or by sexual orientation or race); systems to expand or create facial databases by capturing data indiscriminately through the Internet or audiovisual recordings and television; the recognition of emotions in the workplace and in educational institutions; social scoring (systems that score people based on their social behavior or personal characteristics); systems that manipulate human behavior and AI used to exploit people's vulnerabilities (e.g., age, social or economic status).

Biometric Surveillance

On the other hand, although real-time biometric surveillance systems will be allowed in public spaces, they can only be used by law enforcement and will be very limited and surrounded by strict safeguards: judicial authorization will be required and the list of crimes that authorize it will be very restrictive. In the case of "ex post" use, it will only be allowed for the search of a person convicted or suspected of having committed a serious crime. In the case of real-time surveillance, its use will be limited in "time and location" and will only be allowed for the search for victims of kidnapping, human trafficking or sexual exploitation, for the prevention of a "genuine and foreseeable" or "genuine and present" terrorist threat, i.e. one that is occurring at the same time. or for the location or identification of a suspect in specific crimes: terrorism, trafficking, murder, kidnapping, rape, armed robbery or an environmental crime, among others).

As for the other major point of the negotiations, the regulation of generative artificial intelligence systems on which models such as ChatGPT are based, these will have to meet transparency criteria, such as specifying whether a text, song or photograph has been generated through artificial intelligence, as well as ensuring that the data that has been used to train the systems respects copyright.

This is a technology that was barely known when Brussels proposed the law in April 2021, so measures and safeguards had to be incorporated a posteriori, which was also a reminder that legislation must be adaptable to systems of the future that are still unimaginable. The regulation does not prohibit their use, but it does establish a series of criteria to detect models that may generate a high risk depending on the context in which they are used and obliges their developers to comply with stricter safeguards before bringing them to market.

No other country has yet to be regulated as fully as the European one. U.S. President Joe Biden signed a decree in October requiring technology companies to notify the government of any progress that poses a "serious risk to national security." Days later, British Prime Minister Rishi Sunak convened a summit that resulted in the first commitment of 28 countries and the EU on these systems (Bletchley Declaration) and the creation of a group of experts to monitor their progress.

If no unforeseen events occur, i.e. if neither side backs down and both the states and the European Parliament ratify the law in the coming months (negotiators do not believe that the process can start before February, already under the Belgian European presidency), the AI Act should be able to fully enter into force by the end of 2026, although some parties will do so sooner: the AI Office is expected to start operating as soon as the regulation is ratified, while the ban on prohibited AI systems will reach six months and the requirements for generative AI systems and models will reach 12 months.

You can follow EL PAÍS Tecnología on Facebook and X or sign up here to receive our weekly newsletter.

Source: elparis

All tech articles on 2023-12-09

Similar news:

You may like

News/Politics 2023-12-09T06:59:09.115Z
News/Politics 2024-02-21T10:22:42.136Z

Trends 24h

Latest

© Communities 2019 - Privacy

The information on this site is from external sources that are not under our control.
The inclusion of any links does not necessarily imply a recommendation or endorse the views expressed within them.