Europe will ban artificial intelligence with higher risks

After years of discussions, and when artificial intelligence (AI) has already jumped from fiction to reality and from academic debates to everyday conversations, the European Parliament votes today on the legislative project that, if all goes according to plan, will from 2024 it will become the first Law that will regulate AI in the world. Predictably, and as has happened with data protection, in which Europe was also a pioneer, other countries will later look at the standard to develop their own legislation.

The European standard will divide AI technologies into four risk categories, from minimal to unacceptable. For example, systems that use subliminal techniques or deception to try to manipulate behavior, exploit the vulnerabilities of groups and individuals, or score the social reputation and integrity of citizens will be directly prohibited.. Others, depending on the level of risk, will be subject to audits or will have to clarify what kinds of data they use.. The intention is to “ensure ethical and human-centered development” of this revolutionary industry, sums up the European Parliament.

“It is a complex standard because we are facing one of the technological revolutions in the history of humanity,” says Sergio de Juan-Creix, associate professor at the UOC's Department of Information and Communication Sciences.. «You can imagine all the interests at stake: the public consultation processes, different opinions, the lobbies in Brussels, the irruption in this entire GPT Chat process, which has forced Parliament to include amendments to also regulate this type of generative AI. … In the end, it couldn't be more complicated.”

The point that has been most debated in recent days is that of remote biometric identification: “You are walking down the street, and the cameras are recognizing your face and following you,” describes Juan-Creix.. In principle, this use of AI will be prohibited. However, and as reported by Euractiv, the European People's Party wants to make this point more flexible so that, in case of serious crimes or extreme necessity, and with judicial authorization, this kind of system can be used. Until the text finally becomes Law, this and other debates will continue.

The proposal, which has been working on since 2019 and has already gone through the European Commission and the parliamentary committees on the Internal Market and Civil Rights, will be expectedly approved today, except for surprise. Although, following the complex European procedure, it must now be submitted to new negotiations in the Council, where the member states participate. As indicated by the European Parliament, the objective is to reach a provisional agreement before the end of 2023. Finally, Parliament and the Council will have to formally approve it before it becomes Law.

Meanwhile, technology has advanced rapidly. The clearest example is generative systems, such as ChatGPT. The new standard will force content creators to report if they have used any of these tools, while developers must publish the copyright-protected data they have used to train the system.

«It is about the AI having its ethical parameters, risk, etc., so that different businesses appear on a safe regulatory framework. I think this benefits everyone.. If not, we play in the wild west”, sums up Juan-Creix. «If the regulation is well done, well thought out and is technologically neutral, that is, it allows the survival of a norm despite the technological changes that may exist, it will provide a regulatory framework and legal certainty; and what better way to attract investment than to have legal certainty. When there is legal uncertainty is when there are investment problems”, he contrasts.

The standard has so far gone relatively unnoticed, but it will gain more and more focus as it nears its final form.. Yesterday, Amnesty International expressed its fear that some of the provisions of the current proposal will be forgotten along the way, particularly those that protect against “discriminatory systems.”

For his part, Sam Altman, CEO of ChatGPT, was “optimistic” last week about the possibility of global coordination to regulate this technology, in line with the recent proposal by the British Prime Minister, Rishi Sunak, to host a global summit in autumn. One way or another, we'll be hearing a lot about AI in the coming months.