The new EU AI Act is under threat from lobbyists, experts and the public agree

Lawmakers vote on the Artificial Intelligence act Wednesday, June 14, 2023 at the European Parliament in Strasbourg, eastern France.
Lawmakers vote on the Artificial Intelligence act Wednesday, June 14, 2023 at the European Parliament in Strasbourg, eastern France. Copyright Jean-Francois Badias/Copyright 2023 The AP.
Copyright Jean-Francois Badias/Copyright 2023 The AP.
By Pascale Davies
Share this articleComments
Share this articleClose Button

Foundation and powerful AI models such as ChatGPT and Bard have become a sticking point in the EU’s AI Act.

ADVERTISEMENT

Artificial intelligence (AI) experts and the public have voiced concern that the EU AI Act is under threat from lobbyists trying to remove proposed regulation of powerful AI systems known as foundation models.

Experts signed an open letter calling on European governments to regulate AI rather than companies doing so themselves as the bloc prepares to finalise its flagship bill to regulate AI on December 6.

In the letter addressed to French president Emmanuel Macron, Italy’s prime minister Giorgia Meloni and Germany’s chancellor Olaf Scholz, Atomium, a European science media and democracy NGO, said regulation was under threat due to “pushback” from these governments who are in favour of regulation by the AI companies.

The letter, dated November 26, argues that this was delaying the AI Act’s approval and that if a company self-regulates, it puts human rights at risk as they "may prioritise their profits over public safety and ethical concerns".

Widespread public support for AI regulation

The negotiations, known as the trilogues, appeared to come to a consensus at the end of October.

But foundation and powerful AI models such as ChatGPT and Bard have become a sticking point. In November, negotiations were paused as France, supported by Germany and Italy, opposed binding rules for foundation models.

Meanwhile, more than two-thirds of European citizens support the regulation of foundation models, according to Control AI’s European poll published by YouGov on Wednesday.

More than 80 per cent of the 1,000 respondents in France, Germany, Italy and Spain said they wanted companies creating AI models to be legally liable for damages, instead of shifting the blame to the small and medium-sized businesses distributing or deploying them and users.

Control AI said an anonymous European Parliament official described the move as a "declaration of war" and akin to passing a climate law that excludes oil companies.

Spain, which currently holds the EU presidency and is negotiating on behalf of EU countries, is seeking a revised mandate of the AI Act. The first part of the mandate will be discussed at the Committee of Permanent Representatives on Wednesday.

The part of the mandate on foundation models will be discussed on Friday.

Share this articleComments

You might also like