AI investors will shun Europe if rules are so ‘perfect’ they become a burden, EU lawmaker warns (COPY)

AI investors will shun Europe if rules are so ‘perfect’ they become a burden, EU lawmaker warns

12 September 2023
By Nicholas Wallace and Matthew Newman

As EU legislators work to finalize the world’s first comprehensive regulation on the use of artificial intelligence, a leading European Parliament member is urging colleagues to take a “balanced approach” to ensure investors aren’t put off the European market.

The AI Act needs to strike a balance between protecting people from the risks of AI and making Europe an attractive place to develop the technology, Bulgarian politician Eva Maydell told MLex in an interview.

Dialogue with international partners is crucial to “make sure that we do not isolate ourselves” in pursuit of “the perfect rules,” she said. Otherwise, there’s a danger that the rules could turn out “so burdensome or so uninviting for investors that that the ones that are here will leave, and the ones that potentially can invest here will not prefer Europe.”

Maydell is one of two lawmakers representing the center-right European People’s Party on the European Parliament’s AI Act negotiating team, which is now in talks with EU governments to finalize the new regulation. The other is German lawmaker Axel Voss. The lead negotiators are Italian socialist Brando Benifei and Romanian centrist Dragoş Tudorache.

‘Global standard’

The AI Act will require companies to get AI systems certified by national authorities before putting them on the market for use in “high-risk” scenarios, listed in the law’s annex, such as insurance assessments, school admissions, hiring and firing or the judicial system.

Maydell sees that the AI Act is a good way for the EU to project its regulatory influence around the world, because other jurisdictions may seek to impose their vision of the technology. The parliament’s text, for example, would ban some contentious uses of AI such as for biometric surveillance, emotion recognition and predictive policing AI systems.

“We, as Europeans, along with other international democratic nations, are able to set the standards of the future of AI governance, because my worry is that if we do not do that, there will be other global players, such as China, that will surely do that, but they will do it in a very different manner.”

But Maydell said it remains important that the final text is balanced and “promotes AI development and deployment and that it keeps Europe globally competitive.”

Maydell is also focused on working with other jurisdictions so that the EU’s law is a “global standard.”

“This international cooperation has to be the underlining  principle for the AI act to make any sense whatsoever,” she said. “So it's more about global rules than just us patting ourselves on the back and saying we did a great job.”

Central enforcement

The parliament’s draft of the AI Act, approved in June, would create a European AI Office for more centralized enforcement, something not envisaged in the European Commission’s original proposal in 2021.

The EPP has traditionally been “skeptical” of creating new EU bodies, Maydell said, but “I think we should take in mind the realities of today: The technological development has changed. Its speed, its effects are different compared to five or 10 years ago.” Consequently, “it might make sense that we move toward such an idea.”

Support for more centralized enforcement is gathering momentum, she stressed, particularly as politicians in the US and UK support the idea. “This is also a direction in which the parliament has already pushed, and I think at some point we will probably end up there.”

Even if EU governments agree to more centralized enforcement, national regulators will still be on the AI Act’s front lines. Maydell believes the AI Office will work effectively with EU governments that want to regulate the technology as well.

“For me, it's quite clear that it will be a mix probably, of member states’ authorities, but also central coordinating institutions, whether that's here in Brussels, or its special AI Office,” she said. “Some tasks could be better suited for member states, while others are better suited for Brussels.”

The parliament envisages the AI Office providing guidance and coordinate joint cross-border investigations. This differs from EU governments’ draft of the law, which doesn't include an AI Office, but foresees the creation of “market surveillance authorities” that would oversee AI in financial services, consumer products and law enforcement.

Under the parliament’s approach, governments would be required to establish a national supervisory authority, which would also represent the member state at the management board of the AI Office.

Innovation

Maydell said there are several provisions in the parliament’s text that support innovation, such as the inclusion of “regulatory sandboxes.” These allow companies to develop, train, test and validate AI systems under the direct supervision of national authorities before they are placed on the market or put into service.

The parliament wants to make sandboxes mandatory, while EU governments made them optional in their text, agreed last December.

One of the benefits of sandboxes is their support for innovation, and there have been some victories on this front in the parliament's draft, Maydell said.

“I believe, currently, the text has a number of good provisions, which explicitly promotes innovation in Europe.” For example, she said, “we have an explicit exemption for research and open source. I have also pushed very hard for a new article to be included on regulatory sandboxes, which would allow smaller companies to innovate, to bring their AI products to the market faster.” 

Year-end goal

The parliament and Council of the EU have so far held two rounds of talks to hammer out their differences and reach a consensus on the AI Act. A third round is scheduled for Oct. 2 and Oct. 3.

Maydell sees this next round as critical for achieving a goal of concluding talks this year. “We have been seeing a very strong political will over the past couple of months from the parliament, from the council and from the commission to wrap up the work by the end of the year,” she said.

“We are the first to introduce comprehensive AI legislation, and we would like to see this through,” she said, referring to the need to conclude talks ahead of the legislative agenda being disrupted by campaigning for European elections in June 2024.

But it’s too early to say whether talks can be wrapped this year or will have to extend into 2024, Maydell said. “I might be able to give you a more accurate answer on that in the next month, month and a half … but we indeed still have to tackle some of the more political topics.”

For access to our Data Privacy & Security news and analysis in real time, start your free trial today.