When the Chinese artificial intelligence startup DeepSeek took the world by surprise two weeks ago, one big fear was that it meant China had just leapfrogged American know-how in AI. Another fear, especially among U.S. developers and policy wonks, was that the public might see the moment as proof that massive investments in AI infrastructure were no longer necessary. If cheaper AI platforms work just as well, why sink so much cash into creating expensive frontier models? But this week, all signs point to those concerns being overblown. In fact, ahead of the AI Action Summit in Paris, governments have doubled down on a sovereignty-first AI strategy, with officials publicly rallying behind their AI “national champions” and prioritizing strategies to race ahead on the development of the fast-moving technology as a way to hedge against tech dependency on other countries. U.S. tech companies were already urging Washington to support their rapid computing and energy infrastructure buildout — and if anything, DeepSeek has seemingly given their push a new sense of urgency. It earned AI innovation and competition a spot in the whirlwind news cycle, jumping ahead of talks on AI ethics, discrimination and how to regulate the technology. “What DeepSeek has really done is capture public attention in a way that I haven’t really seen since maybe ChatGPT,” said Oliver Stephenson, the associate director for AI and emerging tech policy at the Federation of American Scientists. “That really boils through into how policymakers are paying attention, and that just shifts the entire ecosystem of Washington, D.C., and policymakers around the world to really focus again on this as a thing that they need to be paying attention to.” Meta, Amazon, Alphabet and Microsoft — the four biggest spenders on data centers — plan to ramp up their expenditures this year, altogether up more than 45 percent from 2024. OpenAI CEO Sam Altman wrote a blog post defending the scaling laws that he says show “you can spend arbitrary amounts of money and get continuous and predictable gains.” The chief of Google Deepmind, Demis Hassabis, said that while its cost-effectiveness is exaggerated, DeepSeek shows “extremely good engineering” and “changes things on a geopolitical scale.” The lean into national sovereignty as a priority in the AI race — rather than safety, fairness, humanitarian issues, or the other big policy considerations around the fast-growing tech — picked up near the end of the Biden administration, which rolled out a rule dividing the world into tiers that determine access to American AI chips for allies and adversaries. In another parting move, Joe Biden released an executive order to fast-track the development of AI infrastructure on federal lands, that in his words, will set “our nation on the path to ensure that future frontier AI can, and will, continue to be built here in the United States.” In Europe, DeepSeek’s nimbleness is viewed by some as a beacon of hope for their own eventual breakthroughs, and leaders are now sharpening their message. “The future of AI is a political issue and an issue of sovereignty and strategic dependence,” French President Emmanuel Macron wrote Sunday, when he unveiled a $100-billion-plus pledge for private AI investment Sunday — France’s answer to President Donald Trump’s Stargate project. He argued that in the U.S.’s race against China, the countries not currently at the forefront but caught in between — like France and India — could find a “third way” to becoming AI powers. “What he seems to be saying is that it's France’s and Europe’s goal to be independent in AI development and deployment from the US and China,” said Pablo Chavez, a tech expert a the Center for a New American Security. It’s also adding a new layer of complexity to the whole conversation about what “sovereign AI” even means. Countries have already been seeking control over critical technologies like AI, and pouring billions into national supercomputers and proprietary models, but traditionally EU member states have thought of tech sovereignty as a target among themselves. AI sovereignty referred to one country’s control of its own AI stack. That appears to be changing: Macron is proposing “a new alliance of countries that would create what's really a third pull of AI development as kind of a balance against and in competition with the U.S. and China,” said Chavez. “It might be of some concern to American policymakers,” he noted. “Vice President Vance is attending the summit, and perhaps that's an opportunity for him to engage on bringing the European effort closer to the U.S.” Some skeptics worry that overly restrictive, nationalistic efforts will crowd out collaboration on other important policy challenges, like managing the potential safety risks and environmental impacts of AI. “I suspect we're going to see increasingly organized backlashes against the expanding energy demands, particularly at local levels,” said Stephenson. “If you are racing to try and develop artificial intelligence, you are going to spend less time focusing on safety, and that is something that I think we should all be pretty concerned about.” Chavez, however, sees the discussion within governments about DeepSeek’s security practices as a chance for global collaboration. Countries from Canada to South Korea have banned the app on government devices, and House lawmakers just introduced a bill doing the same. “There was definitely a common response in the U.S. and across the Atlantic that there were — and are — potential problems with DeepSeek, from a cyber security perspective, from a privacy perspective, from a data access perspective, and those are clearly opportunities to continue to cooperate and engage internationally,” said Chavez. “With AI policy and even on the business side, it’s always a collage of issues, and it’s just a question of what issues are going to bubble up in the moment.”
| 
