To shush AI Act critics, the EU fine-tunes innovation pitch

To shush AI Act critics, the EU fine-tunes innovation pitch

BRUSSELS — The European Union has a message for its artificial intelligence naysayers: It can innovate as well as regulate. The EU was quick out of the gate to regulate AI — it was one of European Commission President Ursula von der Leyen’s key political promises back in 2019. But as the bloc puts the finishing touches on its AI rulebook, some critics worry the new rules could hamper Europe’s competitive position against the United States, China and other rivals. To assuage those fears, some EU executives — including von der Leyen — have started talking up their efforts to give European startups what they need to keep up in the global AI race, such as staff, money, data and computing power. This week, the European Commission is set to present a plan to link the bloc’s AI startups with European supercomputers, addressing the companies’ need for vast amounts of computing power to stay competitive. You may like Earlier Commission agendas had the plan down for February 21, suggesting that the EU executive accelerated its AI innovation efforts — just as EU member countries are deliberating whether to green-light the bloc’s proposed Artificial Intelligence Act. Commission chief von der Leyen gave a sneak peek of the plan to a global audience last week during a talk at the World Economic Forum in Davos, Switzerland. In an address that was starry-eyed about the future of tech, von der Leyen said, “AI is also a very significant opportunity, if used in a responsible way,” calling herself a “tech optimist.” Von der Leyen also singled out an area where Europe should lead the pack: industrial AI. During this mandate, the Commission has already laid the groundwork to improve the sharing of machine-generated data, which could prove to be key for automation and industries’ use of AI. The new plan to provide startups and SMEs access to European supercomputers is similar to what “Microsoft is doing for ChatGPT by running it on its own supercomputers,” von der Leyen added, referring to the corporate front-runners in the global AI race. The innovation push’s timing might not be coincidental, though. EU legislators are in the final stages of their work on the AI Act. Some member countries, especially France, attacked the bill in the lead-up to and aftermath of a political deal that was struck in early December, claiming it could discourage European innovation in the flourishing field. France was concerned that the possible introduction of new compliance rules for so-called foundation models — advanced AI architectures similar to that which powers OpenAI’s ChatGPT — would hamper its national champions, such as open-source AI company Mistral. Germany and Italy also shared these worries. In November last year, Berlin, Paris and Rome circulated a statement calling for looser requirements for foundation models than those envisioned in some versions of the AI law drafted by European policymakers under the Spanish EU Council presidency. The AI Act’s final text, penned in late December under Spain’s stewardship, which POLITICO has seen, largely cleaves to the demands, replacing the concept of foundation models with the wider category of general-purpose AI — as the three countries had requested. Still, France has remained noncommittal on whether it would back the final version of the AI Act. At the same time, the EU’s innovation push isn’t entirely isolated from discussions on how to mitigate AI’s risks — a reminder that, while a focus on competitiveness has long been in the ether, regulatory considerations are always part of the conversation. A provisional header European Innovation Commissioner Iliana Ivanova, who was also present in Davos, told POLITICO that, indeed, a “balanced approach” was needed. “Obviously, the contribution[s] of AI and supercomputing to increase productivity, increase efficiency and bring much higher efficiency gains are very clear,” she said. At the same time, “We have to … look at the ethical aspects [and] the human centricity of these projects.” Gian Volpicelli contributed reporting .