STRASBOURG — Ursula von der Leyen has declared herself an AI “optimist.” Yet in the face of rising warnings of the technology’s threats to humanity, the European Commission president on Wednesday sounded far more like a realist — and one with limited room for maneuver.
Von der Leyen didn’t shy away from the topic in her State of the European Union address, calling AI one of the “two tipping points of our time” and acknowledging the dangers it poses.
"The warnings about the future of AI should make us more serious about safety," said key European Parliament AI lawmaker Axel Voss after the speech, touting von der Leyen's focus on "Europe's human-centered approach to AI."
Advertisement
Still, the Commission president's most concrete response to recent calls by top Silicon Valley bosses to AI development was a cordial invitation to join her in a “discussion” on the matter.
And while citing the EU’s standing AI rules as a possible building block for global regulations, von der Leyen didn’t use her much-anticipated speech to call on fellow world leaders — particularly U.S. President Donald Trump and Chinese President Xi Jinping — to come together to act.
Her diffidence risks turning Europe into a bystander amid a global scramble over calls to regulate frontier AI, which the U.S. leader has dismissed as a “hoax” — just ahead of a high-level summit between Trump and Xi that will likely explore how to set global standards on AI safety.
Von der Leyen identified “the right diagnosis” in her speech, said Max Reddel, CEO of the Arq Foundation, a Brussels-based think tank focused on AI. But, he added, “the question is whether we can move at the speed and decisiveness she calls for.”
In her speech, von der Leyen had to cater to two demands: that Europe stay competitive in the global AI race while also meeting the moment on global AI safety — a frenzy sparked by a warning from a former Anthropic engineer that AI could lead to the end of the human race.
Von der Leyen said she was an “optimist” that AI could benefit humanity, but noted that “risks have sharply come into focus.”
Advertisement
Europe also finds itself in a bind in addressing AI safety because it is playing catch-up on the business and tech fronts, without a real frontier-lab competitor to U.S. and Chinese firms.
Brando Benefei, an Italian MEP, noted that Europe has regulation knowhow but is playing catch-up with the U.S. and China in the race to build out the technology. “Von der Leyen rightfully says that the AI Act positions Europe to lead on AI safety,” he said of the dilemma, while “Europe must [also] now build capacity.”
In her 73-minute address von der Leyen devoted a notable portion to AI risk scenarios, citing an incident this summer in which OpenAI-powered agents hacked another AI company, and issuing warnings regarding “self-recursive improvement,” which is when AI models create a smarter version of themselves, leading to an unstoppable loop.
The Commission president held the line that the EU executive has been repeating for a week in the face of such threats: We have a law for that. Under the bloc’s four-year-old Artificial Intelligence Act, she said, companies like OpenAI and Anthropic must address the risk of losing control of a model.
Von der Leyen largely sidestepped giving her own assessment of whether the development of frontier AI models should be d, citing calls from tech CEOs to do exactly that.
“I will invite the main frontier labs for a discussion on how we can support ongoing industry efforts to pace the frontier,” she said.
Advertisement
But that part of the speech didn’t include an appeal to AI frontrunners U.S. and China, and their leaders Trump and Xi, to act on it — only an offer to middle powers like the U.K. and Canada “and others” to team up on testing the models.
Von der Leyen received good marks from MEPs and AI safety experts for her comments on what is within Brussels’ remit, which highlighted the limits on the actions the EU can take.
“Finally, the Commission is addressing the frontier risks,” said German Greens lawmaker Sergey Lagodinsky in a statement shared with POLITICO. He praised the “middle powers alliance and dialogue with AI labs” as “the right step.”
But he also urged the Commission to “limit the risks” by implementing the AI Act “ambitiously.”
Daniel Privitera, executive director of the Berlin-based KIRA Centre AI think tank, commended von der Leyen for acknowledging the risks, but said more is needed than just a meeting.
“Inviting CEOs to talk about how Europe can help with efforts to pace the frontier is a first step, but Europe can and should do more to help with a safe transition to transformative AI,” he said.
He added that her pitch to team up with the U.K. and Canada to test models is “promising,” as it gives the middle powers “strategic leverage vis-a-vis the two AI superpowers in the world.”
But mentioning the testing of models is one thing; promptly creating a mechanism for doing so is quite another. “We urgently need these things,” Privitera said, “as catastrophic risks from frontier AI become more salient.”