Brussels persists and signs on to the AI regulations. On Monday, three European commissioners took the floor publicly to detail the initiatives they intend to implement even before the entry into force of the future European law on AI.
Unusual, these concurrent declarations illustrate the general mobilization of the EU, while warnings multiply on the serious risks that technology would pose for society.
In reality, the EU wants at all costs to show that important measures, all voluntary for companies, are taken, while the law on AI will not see the light of day for two or three years.
Commissioner Vera Jourova (Values and Transparency) thus announced on Monday that she wanted to ask companies that had signed the EU Code of Practice against online disinformation – Microsoft, Google, Meta, TikTok… but not twitter which caused a sensation when it withdrew from it last week – to propose an “IA label” to identify texts, images, audio and video generated by this technology and to warn users.
Thierry Breton (Internal Market) is working on an “AI Pact” intended to help companies prepare for future EU legislation and ures having convinced Google to participate .
Margrethe Vestager (Competition) is working on an international code of conduct with G7 partners India and Indonesia, in an attempt to harmonize the principles of future legislation throughout the world.
Obama and Merkel sharing ice cream at the beach
Beyond the obvious rivalry between these three Commissioners who spoke separately on this subject, the EU no doubt wants to show that it has taken on the full dimension of the problem, whereas some anticipate, with the rise of AI, a risk potentially as serious as those posed by pandemics or nuclear war.
For Vera Jourova, the fight against disinformation linked to AI thus requires “clear labeling of this type of content”. In particular, the commissioner wants to put an end to fake images like those of Pope Francis in an immaculate Balenciaga down jacket or Barack Obama and Angela Merkel sharing an ice cream on the beach.
She also wants signatories “who integrate generative AI into their services, such as Bing chat for Microsoft, Bard for Google”, to provide “the necessary guarantees so that these services cannot be used by malicious actors to generate disinformation “.
Voluntary, the code of conduct is therefore not mandatory for companies. But similar “labeling” provisions are in the DSA Digital Services Act, which will come into force from next August, and in the future AI Act. Which would therefore force, de facto, Twitter, in particular, to apply it…
Educate the industry
“We are not here to ban, we want to explain what our democracy has decided to put in place to organize our way of living together by mitigating the risks in the digital space”, explained Thierry Breton, qualifying the AI Pact ” antechamber” of the law for AI. “We want to ‘educate’ the industry, help companies prepare to change their ways.”
“We have a sense of urgency. The idea is to impose obligations on the industry itself, the code of conduct will allow us to speed up those of the signatories”, summed up Margrethe Vestager, uring that the boss of ChatGPT, whom she met there a few days ago, had “a positive idea” of this initiative.
Nevertheless, the recent meeting of the Trade and Technology Council between the EU and the United States has shown the limits on how to regulate AI, as the approaches of the continents diverge.
The Biden administration is particularly divided on the future legislation of the EU, one of the most advanced. Next week, the European Parliament must vote to establish its position on this bill, considered, for the time being, as the strictest in the world.