Regulate artificial intelligence (AI). This is what Sam Altman, the CEO of OpenAI, himself claimed when he recently testified before the Senate Judiciary Committee. While he says he’s confident the innovation will benefit people on a large scale, Altman believes lawmakers need to create a framework for AI designers to avoid causing significant harm to the world. Just two weeks later, however, on a triumphant tour of European countries where he met top political leaders, Sam Altman warned that his company could cease operations in Europe if it was unable to comply with the provisions of the new AI legislation that the European Union is preparing. What game is Sam Altman playing?
Generative AIs, classified as “high-risk systems”
In the United States, his initiative was seen as a way to secure OpenAI’s technological lead by reducing the capacity of new entrants. It has been heavily criticized by open source players, who produce many alternative models, but who obviously will not have the ability to comply with the expected regulations. And this is exactly the trap into which Europe is falling. On June 14, 2023, MEPs adopted Parliament’s negotiating position on the AI Regulation. Talks will now begin with the countries of the European Union within the European Council. The aim is to reach an agreement by the end of the year. In this text, generative artificial intelligences are classified as high-risk systems that will require a long and costly certification procedure.
Homologation: better armed American giants
The regulation establishing harmonized rules on artificial intelligence would penalize proprietary software and so-called open source free software equally. Free software developers and hosting services such as GitHub – as importers – would thereby be liable for making unlicensed models available in Europe and face massive fines of €20 million. euros or 4% of worldwide turnover. As the text has an extraterritorial scope, the latter risk prohibiting full access to their site, including models that have nothing to do with artificial intelligence, to avoid any risk.
This regulation will therefore offer an extraordinary opportunity to the American giants who have all the resources to follow these complex certification procedures: that of obtaining a monopoly on the supply of generative AI models to European companies. This will be all the more true as the text explicitly prohibits access via an automatic connector (API) to models based outside Europe for services provided in Europe. European start-ups and SMEs will be exempt from fines. The big deal: in any case, they will no longer have access to the best technologies to develop and offer innovative products and services.
European cacophony between small and large states
Let’s also keep in mind that open source is, today, the only way to serve customers who refuse to have their data sent to the United States. How can you imagine embarking on such a risky exercise with such balls on your feet? And that’s not to mention the instability of the rule, which opens wide areas to the interpretation of the regulators with whom you have to register. Paul Graham, co-founder of the famous incubator Y Combinator and who returned from Silicon Valley to London, declared categorically that if he were a creator of start-ups devoted to artificial intelligence in Europe, he would move.
On technology, finally, the European Union does not speak with one voice. In 2019, the new President of the European Commission, Ursula von der Leyen, pledged to make the 2020s Europe’s “digital decade”. Regulations have followed one another in all segments of technology, from strengthening data protection to artificial intelligence, including content moderation, cybersecurity and even computer chips. But this legislative fury reflects more the concerns of large countries, such as France and Germany, eager for regulations, to the detriment of small states, which are more open to the digital issue.