The EU’s AI Act: technology tamed?

The European Parliament yesterday passed the world’s first law regulating artificial intelligence. The legislation defines different risk levels for different applications. Programmes such as facial recognition software that are considered particularly risky are to be banned; others will only be allowed under certain conditions. While some commentators welcome the decision, others fear overregulation.

La Repubblica (IT) /

Europe on point here

La Repubblica is delighted:

“Europe is known for writing too many rules. Some say this is why innovation takes place elsewhere, in the United States or in China. But with artificial intelligence, the technology that promises to change everything, it is the innovators themselves who are calling for rules. And this time the European clock seems to be ticking fast: yesterday, the parliament in Strasbourg gave the green light for its version of the AI law. ... The final text will now be negotiated with the governments with the specific aim of having it adopted by the end of the EU legislature. This would make Europe the first democratic power to have regulations on AI.”

Filippo Santelli
Frankfurter Allgemeine Zeitung (DE) /

Slamming the brakes too hard

The rules the European Parliament is pushing for are far too strict, criticises the Frankfurter Allgemeine Zeitung:

“The majority of applications are harmless. ... ChatGPT does not pose a threat when it acts as an assistant for Internet searches. It’s a different matter when AIs make decisions that affect people’s well-being, whether it’s autonomous driving or lending money. Then it must be guaranteed that they have been trained using solid, non-discriminatory data. The EU Parliament should have left it at that. But apparently the ChatGPT shock was just too big. The MEPs want generative AIs to be subject to comprehensive checks for risks, regardless of their field of application. This is exactly the kind of slam on the brakes that Europe does not need. The EU urgently needs to step in and readjust the legislative process.”

Hendrik Kafsack
Rzeczpospolita (PL) /

EU’s regulation mania could hinder progress

The new act could set off a damaging cascade of legislation, fears Paweł Rożyński, a journalist at Rzeczpospolita:

“The problem is that the AI Act is just the beginning. The development of artificial intelligence will certainly progress faster than the legislators’ measures. And that calls into question the effectiveness of the regulations. They will have to be constantly updated or completely amended, which will lead to a plethora of new laws. I fear that Europe will have a hard time keeping its urge to regulate under control. And the consequences of over-regulation could be disastrous for the economy, which artificial intelligence will take to a whole new level.”

Paweł Rożyński
RFI România (RO) /

Scope could be insufficient

RFI România questions the effectiveness of the regulations:

“At the bottom of the pyramid are the applications that pose potential risks. Those producing such applications will be obliged to evaluate them and provide the European regulator with a full set of the uploaded algorithms and technical data. This data will flow into a European database that will be accessible to all users. People will be able to find out whether a work to which they own the copyright has been used to produce content with artificial intelligence, for example. ... Of course, one must also ask whether legislation which has only been approved at the European level will have any impact as long as the applications continue to work so well in the rest of the world.”

Ovidiu Nahoi
La Croix (FR) /

No reason to panic

Concerns about AI should not be exaggerated, La Croix believes:

“Before we give in to panic we should remember the worries that accompanied the advent of the digital format, the internet, and of course long before them, computers and even railways. ... It is normal — and ultimately quite human — for the arrival of a new technology that it is believed will revolutionise the way we create, work or talk to each other to provoke anxiety. But this fear is not necessarily bad: it can help us think about the good — and bad — applications of a new technology.”

Anne-Bénédicte Hoffner