A MARTÍNEZ, HOST:
The world's first comprehensive AI legislation is here, and it comes from Europe.
LEILA FADEL, HOST:
European Union lawmakers have approved the bloc's Artificial Intelligence Act. It aims to shape the creation of what they call human-centric AI, with regulations which will apply to any product or service that uses the technology.
MARTÍNEZ: Teri Schultz joins us now from Brussels to tell us what the regulations will and will not do.
Teri, governments around the world are trying to figure out how to regulate AI. So what's the European Union doing?
TERI SCHULTZ: Well, this AI act is pretty ambitious, and the EU is quite proud of it. It took several difficult years of negotiation and public feedback. They had to balance this complicated technology and a very powerful industry with consumer protections and then get 27 governments to agree to the same set of rules. Here's one of the parliamentarians who led work on the law, Dragos Tudorache.
(SOUNDBITE OF ARCHIVED RECORDING)
DRAGOS TUDORACHE: We have a duty to recognize this potential because it is going to be the technology that will be driving us into the future. And at the same time, we realize that, clearly, there are risks. There are concerns.
MARTÍNEZ: All right. So that's the philosophical approach to it, Teri. So how are they going to do it practically?
SCHULTZ: So the AI act will classify systems or applications into four categories according to what's considered their potential risk to society - from low to limited to high, up to unacceptable. And each of these will be subject to different levels of regulatory control. At the low end, there won't be any requirements or regulation. And at the unacceptable end, they'll be banned altogether in the EU.
MARTÍNEZ: So OK, give us some examples of what we're talking about here. I mean, who gets to decide how risky a certain use of AI would be?
SCHULTZ: Well, at the low-risk end are applications such as AI in video games or spam filters - we're all familiar with those. And at the unacceptable end are uses such as biometric identification, which will be prohibited in public spaces, except in special cases, like law enforcement or counterterrorism. But authorities will have to get special permission, and it will be really strictly limited. EU governments are the ones responsible for evaluating and enforcing the act on their own territory, but the bloc has set up an office here in Brussels to help coordinate and make sure it's being done in a fair way across the EU. It'll also provide guidance for companies and developers so they're in compliance.
MARTÍNEZ: Ah, developers - that's the other party in this. What's been the response from the tech industry from those developers, and what happens if companies just don't want to comply?
SCHULTZ: Parliamentarians said it was one of the most heavily lobbied pieces of legislation in years, but they feel like they've both protected consumers and the technology's potential. This is how it was described by Brando Benifei, one of the co-leaders of the parliamentary work.
(SOUNDBITE OF ARCHIVED RECORDING)
BRANDO BENIFEI: Too many citizens in Europe are skeptical of the use of AI, and I think this is a competitive disadvantage. And this would stifle innovation. Instead, we want our citizens to know that, thanks to our rules, they can trust the businesses that will develop AI in Europe.
SCHULTZ: So A, I'm not sure if tech companies are buying that explanation of how more regulation is going to be good for them, but they won't have a choice. The act will get final approval from EU leaders in the next month or two. It'll be gradually implemented over the next two years, and then violators can be fined up to 7% of a company's annual turnover.
MARTÍNEZ: All right. We'll see how it works out. That's Teri Schultz in Brussels. Teri, thanks.
SCHULTZ: Thanks, A. Transcript provided by NPR, Copyright NPR.
NPR transcripts are created on a rush deadline by an NPR contractor. This text may not be in its final form and may be updated or revised in the future. Accuracy and availability may vary. The authoritative record of NPR’s programming is the audio record.
