The European Union took the first step yesterday to achieve legislation for the use of artificial intelligence that places the focus on the protection of human rights. The text coordinated by the rapporteurs Drago? Tudorache and Bruno Benifei maintained their ambitious approach to technology control, especially with real-time facial recognition surveillance in public spaces, which the European People’s Party (EPP) wanted to amend, without success, to establish exceptions such as the disappearance of children or crimes of terrorism.
The approval of the proposal, which received 499 votes in favour, 28 against and 93 abstentions, opens an intense three-party negotiation since yesterday, between the chamber itself, the Council -in which the governments participate- and the European Commission. These last two institutions will raise the application of police exceptions to restrictions such as facial identification.
The guaranteeing nature of the legislative proposal in the protection of privacy contrasts with the requests of security forces of the member countries. The president of the European Parliament, Roberta Metsola, assured that “looking ahead, we are going to need constant and clear borders and limits for artificial intelligence.” “And here’s something we won’t compromise on,” she added. Every time technology advances, it must go hand in hand with our fundamental rights and democratic values.”
The negotiation will have as its main points of friction aspects such as the use of AI for police control, but the deputies who have worked on the proposal know that they now have a very considerable voting mandate. “Today’s result gives us an even stronger position,” said Bruno Benifei, one of the text’s speakers. “It is clear that Parliament does not want us to back down on issues as important as avoiding mass surveillance,” he added.
“We have seriously considered the interests of society and our citizens in terms of privacy. And that is why we have gone a step further than the commission by removing the exclusions for law enforcement, but without stopping thinking about security,” said Drago? Tudorache, the other speaker. The parliamentarian pointed out that the text includes “that biometric identification can continue to be used with due judicial authorization right after”, and not in real time.
Tudorache recalled that the request made by governments and the PPE is already covered by current legislation. “If there is a terrorist attack on the street right now or if the police have received a notice that a child has been lost, images can already be obtained from surveillance cameras, in accordance with current laws and with judicial authorization, biometric identification can be used”, observed the deputy.
One of the innovative elements of the proposal is that the law establishes obligations for AI operators based on their degree of risk. Thus, any system that can manipulate the behavior of vulnerable groups such as children, or those that elaborate a “classification of people based on their behavior, socioeconomic status or personal characteristics” will be considered “unacceptable”.
Within the “high risk” AI systems are considered those that affect security and fundamental rights. Generative AIs, like ChatGPT’s, will need to disclose content that has been generated with their technology, prevent illegal content, and publish copyright data. The Commission hopes to reach an agreement for the artificial intelligence law by the end of the year. The entry into force, however, will not come until at least the beginning of 2026. The first proposal to regulate AI came out of the executive in 2021.
