Europe scrambles for sovereign AI after US cuts access to Anthropic's models

As the US and China surge ahead, is Europe sleepwalking into AI disaster?

A burgeoning genre of fictional AI doomsday scenarios says lagging behind on the technology could threaten the continent's sovereignty It's 2031 and the US and China are about to tear Europe into pieces. The US ploughed vast sums into datacentres and the EU did not. China built robots and Europe did not. American companies "restructured" their workflows around AI and fired people, while EU workers went on long lunch breaks and handed over administrative tasks to the AI model Claude. Now the chickens are coming home to roost. Europe's economy is a shambles because it does not have its own AI. Populism is surging, the euro is wobbling, cyber-attacks are shredding EU businesses. Brexit seemed like a good idea. It looks like the end of the European Union. That, at least, is the vision of a speculative thought experiment, called Europe 2031, penned by Brussels-based thinktankers and published fortuitously one day before the Trump administration decided to block "foreign nationals" from using a much-hyped AI model built by Anthropic, called Fable. In the heady week of G7 talks that followed, the scenario has gone viral - feeding a feverish discussion of the urgency for EU tech sovereignty. It has been read by members of the European parliament and, say its authors, was brought up in track 1.5 discussions between British and German officials earlier this week. Its authors say they feel "vindicated", by the attention it has received and by the fact that one of their predictions - that the US would restrict global access to advanced AI models - appears to have briefly come true. They hope the scenario will spur Europe towards a dramatic course-correction on AI. The piece is part of a burgeoning genre of fictional AI doomsday scenarios, created by obscure figures, which have gained surprising traction among policymakers over the past year. In 2025 there was AI 2027, a thought experiment which culminates in a superintelligent AI killing all of humanity to make way for more datacentres; in February, another speculative scenario imagined AI upending the US economy. (The first was read by US vice-president JD Vance, the second contributed to a stock market wobble.) One complication of all this might be that their thought experiment is at times based on current developments in AI whose outcome is uncertain or in doubt. Maximilian Negele contributed to Europe 2031, he says, because of the "incredible translation barrier" between Brussels and San Francisco, where AI is being developed. Formerly at US thinktank Rand, he left his job this year to focus on the project. "As somebody who travels to San Francisco quite a bit and talks to people there, what is happening in Europe just seemed like a slow-moving car crash to me," he says. The scenario unfolds from the perspective of a fictional bright-eyed Brussels staffer, Caroline Dubois, who has a German friend, Christian Vogt, with a startup in San Francisco. On a visit, she's impressed by America's "70 or 80-hour" working weeks and discomfited by the conviction among tech bros that everything is about to change. Back in Europe, she works to evangelise her well-meaning bosses about the impending AI future - but fails to convince. There's too much scepticism, and most people think AI is a bubble. Things go from there. The Americans spend huge sums on a massive AI building programme - the scenario highlights a real-life $100bn (£75bn) deal between OpenAI and Nvidia, the $300bn agreement between OpenAI and Oracle, and "bulldozers" breaking earth in Texas for an AI datacentre. Europeans, meanwhile, put forward a tepid investment package and ignore advisers' pleas for "a full regulatory carte blanche for datacentre providers". In a matter of years, America monopolises 70% of the world's "compute" - the semiconductor chips that fill the datacentres that power AI models. Europe's economy is meanwhile gasping for air, mostly because its companies have not adopted AI. As AI-powered cyber-attacks shred European firms and unemployment surges, EU officials scramble to parlay their one last bargaining chip - the Dutch lithography firm ASML, which is vital to the production of AI semiconductors - into concessions from Beijing or Washington. But it's too late. The US deploys powerful "frontier AI" spyware and learns the deepest fears of EU officials and also which of them are having affairs. Curtains drop. Christian and Caroline exeunt stage left for a drink. Disaster impends. Sceptical readers might point out that a number of the eye-popping sums and big projects that the authors name-check in describing America's AI ascent have already fallen apart. The $100bn agreement between OpenAI and Nvidia, the biggest AI deal of last year, evaporated in February. The $300bn between OpenAI and Oracle seems doubtful, especially as recent reports indicate the maker of ChatGPT is still billions of dollars underwater as it burns money on datacentre infrastructure. The bulldozers on the ground in Texas may not be bulldozing very much any more, as OpenAI pulled out of the flagship AI project to which that moment in the scenario seems to refer. The authors are sanguine about these matters. Throughout the piece, they pre-empt potential objections - such as AI being overhyped - by suggesting that the hapless European officials have these worries, too, and they end up tragically wrong. "I wouldn't rule out that there's some exuberance and that one or two AI companies might go bankrupt," says Negele. "But what we wanted to get across is a general feel for a version of what we think will happen." He and his co-author, Alex Petropolous, agree that there could be some bumps in the road - including mounting resistance to datacentres in the US. "I mean, people hate AI in general. A lot of people do. People hate datacentres. They destroy the landscape. They support big tech. It's a very, very unpopular policy." The authors of Europe 2031 think that the solution to this is datacentres. Europe needs to build more, faster, ideally in AI zones where matters such as power and planning can be streamlined and deregulated. "I think our view is that the total datacentre supply is quite an inelastic supply. So there will only be a limited number of datacentres built in the world built every year, and the question is, how many of those do you want built in the US? How many of those do you want built in Europe?" says Petropolous. It is further worth noting that the main organisation behind the Europe 2031 scenario, Arq Foundation, based in Brussels, describes itself as "neither an advocacy NGO nor a venture-backed startup" and does not disclose who funds it. Brussels politicians who read it, though, may take away a simpler message: the scenario has crystallised a conversation about the need for Europe to have technological sovereignty. "This scenario, Europe 2031, I believe that some of the parts they mentioned can happen," says Nicolás Casares, a member of the European parliament from Spain. "But I think they are increasing - a bit - the alarms in order to call our attention." The US cutting off Europe's access to Fable, he says, means that the EU needs to ask itself harder questions about who is building its AI infrastructure and who will benefit from it. "What is the added value of having OpenAI or Anthropic datacentres in Europe?" he says. "We are buying a narrative that we need a lot of datacentres not to lose the race for AI. But this is crazy ... we are paving the way for infrastructure that they will use and sometimes not allow us the possibility of using it."

Renting AI from foreign providers is a national security risk, warns Cohere CEO | Fortune

Aidan Gomez, CEO of Canadian AI company Cohere, has spent years planning for the possibility that foreign governments might abruptly cut off access to frontier AI models. Earlier this month, when the U.S. government moved to restrict foreign access to Anthropic's Mythos, it was a striking example of exactly what he'd been warning about. "I think it's woken everyone up to the reality, which is that centralized dependence on a single entity is a structural risk," he told Fortune. "You can absolutely just have access revoked...and services shut down." The U.S. has largely led the global AI race since it began, controls vast amounts of the world's AI computing power, and is home to the majority of leading frontier labs. China is its closest rival, and together the pair hold 90% of the world's computing power, according to data from AI research firm Epoch AI. Outside the two nations, the field of frontier AI developers is more narrow: Cohere, founded in Toronto in 2019, and Paris-based Mistral are among the only prominent companies building and deploying large-scale models from outside the U.S.-China axis. As AI becomes more deeply embedded in critical systems, such as hospitals, financial markets, military logistics, and public services, the question of who controls that technology and who can switch it off is becoming a matter of national strategy. Gomez argues that democracies need to stop renting critical AI from a handful of foreign providers and instead build sovereign systems they can control end‑to‑end. In practice, that means owning or have some control over the full chain of components an AI system depends on: the data centers, the chips inside them, and the models themselves, rather than paying another country's company for access to those things via an API (the software interface that lets you tap into someone else's model over the internet). "This sentiment of renting AI from someone rather than owning it is a national security risk," Gomez said. "You need to fully control it." Cohere already sells its own models in a way that lets governments and companies run them entirely on their own infrastructure, Gomez said. Rather than sending data to Cohere's servers, customers receive the model and the software needed to deploy it inside their own data centers, including on systems that are physically disconnected from the internet and to which Cohere has no remote access and no ability to revoke service. Sovereign AI "Sovereign AI" has become something of a catch-all term, often used to describe many different levels of independence. Some use it to mean controlling the data an AI system is trained on or processes; others mean the ability to build and train new frontier models from scratch; then some mean simply ensuring that sensitive information never leaves a country's borders. The lack of a settled definition has become part of the political debate and can allow governments and companies room to claim progress on sovereignty without necessarily committing to any of its more demanding forms. For Gomez, the central part of sovereignty starts with physical infrastructure. "Domestically controlled infrastructure is the first piece of the puzzle," he said. "Each country needs an infrastructure champion." Only when governments have a domestic player that actually controls the chips, power, and data centers a country's AI systems run on, can they be sure that no foreign firm -- or foreign regulator -- can unilaterally pull the plug, he said. However, the dream of fully sovereign AI inevitably runs up against hard constraints around compute, capital, and energy. Building and running frontier‑scale systems is extraordinarily expensive and resource‑intensive, and most countries will never be able to own every layer of the stack outright. Realistic sovereignty, Gomez said, has to be pursued through alliances. Rather than every government trying to replicate the whole ecosystem at home, Gomez wants democracies to pool their bets around a small number of "champions" and share capacity through trusted partnerships. "It's not the case that every country is going to have each layer of the AI stack inside their own country...instead they need to find multiple partners for those layers," he said. "If we try to spread our resources, it just won't work. We have to back champions and not spread our bets." Cohere has been attempting to partner with other "middle power" nations and companies. In April, Cohere announced a deal to acquire Germany's Aleph Alpha, with the aim of creating a transatlantic AI company anchored in both countries. Two months prior to the deal, Canada and Germany had signed a Joint Declaration of Intent on Artificial Intelligence and launched a sovereign technology alliance. The Cohere-Aleph Alpha deal was positioned as a way to put that political commitment into commercial practice. The combined company, which would retain the Cohere name and be led by Gomez, will be valued at around $20 billion, according to the Financial Times. At last week's G7 discussions, Gomez said leaders and AI executives largely agreed on the need for democratic countries to coordinate on standards and regulation. Now, he wants them to go further and invest in shared capacity, so their critical infrastructure isn't dependent on a few U.S. or Chinese firms. The politics around AI, he argues, are not that different from the technology itself. "Centralized dependence on a single entity is a structural risk; we need to avoid these single points of failure," Gomez said. "I'm a computer scientist, and that's one of the first things you learn: don't build a system with a single point of failure."

Europe risks 'total irrelevance' without sovereign tech: Cohere chief

Paris (France) (AFP) - Europe risks "total technological irrelevance" if it fails to develop sovereign capabilities, the head of Canadian AI startup Cohere warned Thursday on the sidelines of the Vivatech trade fair in Paris. Aidan Gomez spoke to AFP the day after a G7 meeting in Evian, eastern France, where he and other AI luminaries including OpenAI's Sam Altman and Anthropic's Dario Amodei were invited to join leaders of the world's top industrial nations. Without tech built on home soil, "your data, your productivity, your critical industries, your ability to defend yourself, it will be what others allow," Gomez said. Vivatech has been overshadowed by European tech sovereignty concerns, after Anthropic withdrew access entirely to its most powerful models, Mythos and Fable, in response to a White House order that they be kept out of foreign hands. - 'Cold water' - "It's like everyone got cold water splashed on their face," Gomez said. He has striven to place Cohere as a sovereign alternative in both the European and Asian markets. The 2019-founded company builds AI models for businesses and governments, shunning the consumer applications explored by its giant American rivals. Instead, its offerings focus on concrete applications in health, finance, energy and national security. Truly sovereign capability requires "a set of technical guarantees," Gomez said: "infrastructure that you control, that is run by a sovereign entity... meaning a company and champion within your nation." As global tech firms scramble to respond to governments' urgent quest to escape US or Chinese reach, many are offering solutions of their own marketed as "sovereign". But "there's so much sovereignty washing, where solutions that are no at all contributing to diversification or resilience are being painted in that light," Gomez said. Cohere does not operate its own data centres, planning instead to deploy its software on infrastructure owned by local companies like France's Mistral or Germany's Deutsche Telekom. Those AI models "require enormous resources" to develop, Gomez said. "Each country can't build its own model, that's the reality we have to face." "What we need are a number of champions distributed across democracies and then democracies collaborating together to build those champions up," he urged. European ambitions Cohere sees Europe as a natural sphere of expansion in that light, as a "democratic stronghold" against the tech prowess of autocracies such as China. The Canadian company last month announced it was buying German startup Aleph Alpha with backing from both governments. With dual headquarters in Toronto and Berlin, the merged entity is valued at around $20 billion. Since then it has also snapped up Canadian-German biomedical startup Reliant AI. And it signed partnerships in May with Spanish firms Indra, a defence specialist, and Multiverse computing, which develops AI software. Canada and Europe have a "historic alignment (and) values alignment," said Gomez, whose mother is British and father is Spanish. He himself lives in London, where Cohere recently beefed up its offices. The company now numbers around 700 staff and is preparing a new fundraising round. Europe is Cohere's second-largest region for revenue behind North America, Gomez said. "We're seeing a rapid uptake" of AI in Europe, the CEO said. He added that he puts no stock in calls for a pause in AI development to properly weigh the technology's dangers, issued in early June by American giants including OpenAI and Anthropic. "I think no-one is going to pause. It's a very silly posture. It's not realistic," he said. "If anything, Canada and Europe have to go much, much faster."

France and Germany call for European AI sovereignty at VivaTech

When the US suspended access to Anthropic's latest AI models last week, it reminded Europe that rules can change overnight. At VivaTech in Paris, France and Germany said the answer is sovereignty -- and the window to build it is now. The French and German economy and digital ministers issued a joint call to action at the opening of the VivaTech conference in Paris on Wednesday, arguing that Europe must move quickly to build genuine technological sovereignty or risk being left behind as a spectator in the decade's defining technological shift. France's minister for the economy, finance, industry and energy sovereignty, Roland Lescure, opened with an honest acknowledgement of the mood in the room. "I can see some of us are worried [about the risks of artificial intelligence] -- and worry is good. Concern is good," he said, though he cautioned against fear as a guide. "None of us is going to do it alone." Drawing on the historic role of the Franco-German relationship, he argued that when the two nations align, "Europe moves". "I am convinced that France and Germany can make it -- and be at the heart of what comes next. The next 10 years will be important," he said. Germany's Federal Minister for Digital Transformation and Government Modernisation, Karsten Wildberger, addressed the United States' decision last week to ban Anthropic's latest artificial intelligence models for foreign nationals, including the company's employees. "The suspension of access to the most advanced models makes one thing clear to everyone," he said. "This is no longer an access debate; rules can change overnight, and sovereignty means we can still act if things like that happen," he said. Wildberger then said AI sovereignty is not protectionism but a necessity for agency. "Sovereignty is not isolation. It is openness from a position of strength," added. As for Germany, he said the country is implementing a national data centre strategy with a target of quadrupling AI capacity by 2030. The country is also working on a sovereign cloud infrastructure. But he said that government alone cannot carry the effort. "The heroes here are the great startups," he said. "Europe can do innovation, and we do scale, and we do matter." Both ministers pointed to new bilateral initiatives between France and Germany as the practical starting point. "When we pull our technologies together, we shape what comes next," the German minister said. "We will not be spectators of the years ahead." Europe does not lack ideas, or talent, or companies willing to build, he added. "It is not a question of whether Europe is ready. What Europe needs now is courage, ambition, and the discipline to execute and turn this into scale."

Anthropic shutdown ignites calls for sovereign AI across Europe | Fortune

On Friday, when the U.S. government pulled the plug on global access to Anthropic's most powerful AI models, it confirmed some of Europe's worst fears. For the first time, Washington had effectively used what some had dubbed a "kill switch" -- the ability to cut off foreign access to U.S. AI systems. Politicians around the world responded with panic and anxiety about their own countries' dependence on U.S. AI technology. In Europe especially, it reignited calls for what officials describe as "sovereign AI" -- the idea that countries should control the AI models, computing infrastructure, and data that underpin the increasingly critical technology, rather than depending on systems that can be restricted or withdrawn by foreign governments. Europe has, for some time, been heavily reliant on the U.S. for its technological infrastructure. The EU relies on non-EU countries for more than 80% of its technology and 70% of its cloud computing, according to the European Commission and the European Parliament. The U.S. and China together control roughly 90% of global AI computing infrastructure, according to AI research firm Epoch AI, leaving Europe with limited capacity to train or run AI models independently. Over the past few years, and particularly since U.S. President Donald Trump began his second term in office, countries have increasingly seen American behavior as erratic and threatening to the established global order, with everything from unexpected tariffs to demands that Denmark surrender Greenland to the U.S. In light of this, that dependency on U.S. technology has started to feel like a strategic vulnerability. Some governments had begun taking concrete steps to reduce their reliance on the US: the European Parliament replaced Google with Qwant, a French privacy-focused search engine, citing concerns over personal data collection, while Germany, France, and the Netherlands have moved Microsoft off public-sector infrastructure. But the Anthropic shutdown makes the stakes, when it comes to the AI race, harder to ignore. Aside from France's Mistral, whose models currently lag their American counterparts, Europe does not have a frontier AI company, and the continent's data center build-out has also been plagued by high energy costs, capital constraints, and regulation. In the UK, Member of Parliament Al Carns said British hospitals, companies, and researchers had been suddenly stripped of access to Fable 5. Tom Tugendhat, a former UK security minister, said that the incident demonstrated that national security is now more about "code than cannons," criticizing the UK's regulatory approach for prioritizing safety over building competitive AI capacity. Kanishka Narayan, the UK's minister for AI, said the move had significant implications for defense. "We're seeing again and again the most capable models used in drones, counter-drone defence systems and cyber security -- those are now the fundamental fault line for who wins and who loses in warfare," he said. "The central question for our national security and defence is a question of our AI capability." Politicians across the continent also called for Europe to accelerate the development of its own AI models. Many reactions focused on Mistral, the EU's most credible AI contender, which is reportedly in talks to raise $3.5 billion in funding at a $23.2 billion valuation. Former French prime minister Édouard Philippe said the incident showed that AI is now critical infrastructure, as essential as electricity or the internet, and that infrastructure controlled by others is infrastructure that others can unplug. Bruno Retailleau, a French 2027 presidential candidate, said the move should serve as a "wake-up call," arguing that a nation that depends on others for its technology is a nation that can be unplugged overnight. In Finland, MEP Aura Salla said Europe cannot keep building its tech stack on access that can be switched off overnight by a foreign government, while in Germany, MEP Sergey Lagodinsky pushed for the EU to team up with middle powers such as Canada, Australia, Singapore, and the U.K. to pool compute capacity through a consortium approach. "This is something that we have always known, but never really felt," Sandra Wachter, a professor of technology and regulation at the University of Oxford, told Fortune. "Everybody is aware that we are very dependent on technology from the U.S...but we have never quite felt what it's meant to be on the shorter side of the stick -- on the side that actually has to face the kill switch." Europe's policy response Just weeks before the shutdown, the European Commission had unveiled what it billed as its answer to this vulnerability. On June 3, it released the European Technological Sovereignty Package, targeting cloud computing, AI, semiconductors, and open source. Its two main legislative proposals -- Chips Act 2.0 and the Cloud and AI Development Act -- set out objectives for the EU's semiconductor industry and local cloud and AI providers. The Cloud and AI Development Act aims to triple data center capacity in the EU over the next five to seven years. "We cannot afford to depend on others for the technologies that keep our hospitals running, our energy grids stable and our services secure," European Commission president Ursula von der Leyen said. The package targets an estimated €422 billion ($490 billion) in total investment across semiconductors, data centers, cloud, AI, and open-source software over the next decade, though the Commission has yet to confirm how exactly that will be financed, saying it will consult with member states and the European Investment Bank on funding mechanisms. Critics have questioned Europe's capacity to deliver on the commitment. The Commission has deferred to member states on how procurement rules would actually be enforced, and under the proposed regulation, only around 10% of cloud contracts would carry a strong European sovereignty standard, with the remaining 90% open to all suppliers. Competition economist Cristina Caffarra, founder of the Eurostack Industry Initiative, told The Parliment the package was "very feeble," arguing the Commission had watered down previously stronger mechanism for buying European under pressure from Washington. Some experts, such as Wachter, argued that Europe should focus on developing smaller, more efficient AI models, which can be run using smaller data centers and consume less electricity, rather than trying to match the scale of American frontier systems. If the capability gap amounts to only a few percentage points, Wachter questioned whether closing that gap justified the costs of an all-out race. "This whole idea of racing -- if we're not racing now, we're going to be left behind -- is also a narrative that plays into the hands of the people that are developing those technologies," she said. She also suggested that building alliances with like-minded countries around the world offered a more sustainable path for Europe than trying to out-build the U.S. alone. American soft-power The concerns about over-dependence on America extended well beyond Europe. At February's AI Impact Summit in New Delhi, conversations about the sovereign AI issue -- and a potential coalition of "middle powers" -- were front and center. Yoshua Bengio, a Canadian computer scientist and AI pioneer, previously told Fortune the question of AI sovereignty was also one of global power. "It's a question of democracy and a kind of equitable world order in which no one country can use technology to dominate the others," he said. "We don't want to end up in a world where we have two hegemons who each control part of the world." While the U.S. may have had legitimate national security grounds for acting, the manner of the intervention risks undermining America's influence globally, according to Jonathan Iwry, a fellow at the Wharton Accountable AI Lab. "America's global influence rests on more than economic and military power alone," Iwry told Fortune. "It has also depended on moral leadership and being viewed as a reliable steward of an open international order -- one whose institutions, technologies, and markets set the standard and inspire other nations to choose democracy over autocracy." The move may also hurt a broader strategy the U.S. has been pursuing to shape the global AI landscape before China can. Pax Silica, for example, is a U.S.-led technology alliance designed to build secure semiconductor supply chains and advanced manufacturing networks among strategic partners, with the explicit aim of countering Chinese AI development. The alliance -- which includes Japan, South Korea, the U.K., and Israel -- is Washington's attempt to lock allies into the U.S.-aligned AI infrastructure ecosystem. At the New Delhi summit, India became its latest member, a significant diplomatic win for Washington. Frontier AI has itself become an instrument of American soft power, Iwry said, and the way the U.S. exercises control over it will shape whether allies continue to orient their technological ecosystems around the United States. "Exercising unilateral control might increase U.S. leverage in the short term, but it could also encourage allies to reduce their dependence on American AI over the longer term," he said. "The broader challenge isn't just protecting U.S. national security, but doing so in a way that preserves our allies' trust and confidence. Otherwise, the very actions the U.S. is taking to preserve its global leadership might encourage its allies to turn elsewhere."

Sovereignty fears to dog AI enthusiasm at France's Vivatech

Paris (France) (AFP) - Europe's biggest tech trade fair Vivatech opens Wednesday in Paris for a 10th-anniversary edition where enthusiasm for generative AI will rub shoulders with anxiety about the continent's technological dependence. Organisers have recruited Amazon and Blue Origin founder Jeff Bezos as Wednesday's star guest, with French President Emmanuel Macron and Indian leader Narendra Modi expected Thursday. Also joining the roughly 15,000 start-ups showing off their inventions until Saturday will be French researcher Yann LeCun, who made waves this year with a new company focused on physical AI after leaving Facebook's parent company Meta. Meanwhile debate around Europe's tech sovereignty was stoked on the eve of the event by French Prime Minister Sebastian Lecornu. He announced that the country's domestic intelligence agency would break with the American data sifting giant Palantir in favour of local firm ChapsVision. Washington had already pushed tech dependency to the top of the agenda last week. It banned non-American users from the powerful AI models Fable and Mythos, whose creator Anthropic withdrew access for all users in response. European sovereignty With 200 German start-ups in Paris as this year's guest nation, organisers want to make the 10th year of Vivatech a celebration of European cross-border cooperation. "In a time of growing technological and global fragmentation, this spotlight underscores Europe's ambition to affirm its sovereignty and take the lead in innovation," they said. ASML, the Dutch firm that makes ultraviolet lithography machines indispensable to advanced chipmaking, is sending its chief executive Christophe Fouquet to Paris. The company's commanding place in the global semiconductor value chain has recently propelled it to Europe's biggest company by market value. Dancing robots Bringing together founders, investors, industry associations and national delegations, Vivatech is a hive of activity every year. Organisers have expanded the 2026 event to 70,000 square metres (750,000 square feet) from its usual 50,000, and visitor numbers could top last year's record of 180,000. This year tech firms were able to offer demos of their wares for the first time on Paris' Champs-Elysees avenue. Spectacular robotics displays are scheduled for one of the main stages on Wednesday. Chinese firms Unitree and AgiBot aim to again wow spectators with their dextrous humanoid robots, while European start-ups including Genesis and PAL Robotics will also be on show. Fans of AI automation may be more excited by the Thursday appearance of Peter Steinberger, the Austrian creator of the open-source AI agent OpenClaw that took Silicon Valley by storm earlier this year.

Europe frets about US AI as tech world flocks to France for G7, VivaTech

European leaders and tech executives are focusing on technological sovereignty. Discussions at G7 and VivaTech highlight concerns about American AI dominance. Europe aims to build its own AI champions. This involves boosting regional cloud and chip industries. The goal is to reduce reliance on U.S. technology. France is pushing for domestic solutions in government services. Europe's quest for technological sovereignty will dominate discussions at the G7 in France and the VivaTech conference in Paris this week, as policymakers and technology executives fret about American AI, with alternatives remaining scarce. The gatherings come days after the United States tightened restrictions on Anthropic's most advanced AI models for foreign nationals, underscoring Europe's vulnerability that political whims could derail its race to build domestic AI champions. "Tech sovereignty will be top of mind this week at VivaTech," Ana Paula Assis, senior vice president at IBM, told Reuters. "For European organisations to get ⁠this right, ⁠it is vital to understand sovereignty is about having control where it matters - not where the technology is from." The debate reflects a broader dilemma facing Europe: how to maintain strategic autonomy while remaining dependent on American technology companies that dominate cloud computing, semiconductor design, and cutting-edge AI research. The Group of Seven (G7) nations have gathered in Evian, France, where they are meeting top executives from the biggest AI companies including Anthropic, OpenAI, Alphabet's Google, and Mistral to discuss AI competitiveness, regulation, and reliance on China for critical minerals. In Paris, over 180,000 visitors, startups, investors, policymakers and executives including Amazon's Jeff Bezos are expected to attend VivaTech, where ⁠discussions are likely to focus as much on geopolitics and policy as on the actual tech. French startup Mistral, seen as Europe's leading AI contender, is doubling down on partnerships with European firms, particularly in industries where the region says ⁠it has an edge. Despite billions of euros of investment, European AI firms continue to rely heavily on U.S.-controlled cloud infrastructure, chips, and foundational AI models. Europe pushes to boost regional cloud, chips France has been a prominent supporter of European tech sovereignty, with the government increasingly looking to replace US providers in government services. "We cannot rely on tools developed by foreign powers. France must have its own tools," said Prime Minister Sebastien Lecornu on Tuesday. The European Commission is assessing the practical implications of the U.S. export control directive and that measures should not be discriminatory against partners. European policymakers have increasingly framed AI as a matter of economic and national security. The European Commission recently unveiled plans for AI "gigafactories" and large-scale computing infrastructure designed to provide the region with sovereign access to computing power. It has proposed laws to boost domestic cloud, AI and semiconductor industries and cut reliance on U.S. Big Tech, though critics say Europe remains years behind U.S. ⁠rivals. "It's patently clear, if it wasn't before, how important it is for Europe to have access to an AI service that it can control, that will never be switched off on a whim," said telecoms firm Orange in a statement. For companies, boosting sovereignty comes at a cost, however, and they need to weigh that against a perceived risk, said Capgemini chief operating officer Karine Brunet, adding that European cloud alternatives require paying premiums of up to 40%. "The alternative is not simply replacing one provider with another," said Francois Bitouzet, managing director of VivaTech. "It is about building more resilient technology strategies, where companies can draw on European innovation for the most critical parts of their stack while still working with global partners where it makes sense."

AI sovereignty is about options, not ownership - The Korea Times

ÉVIAN -- The U.S. government's sudden decision, on June 12, to restrict foreign access to some of Anthropic's most advanced models is further confirmation that AI is now a geopolitical issue of the highest order. Until recently, countries competed by building services, infrastructure, and applications on top of frontier artificial intelligence (AI) systems. Now, access to the systems themselves is a strategic concern. The prevailing assumption used to be that AI would follow the logic of globalization. Countries would rely on a handful of frontier models, mostly developed in the United States, while competing in downstream services, semiconductors, data, and applications. Access to the most advanced AI systems was largely taken for granted. But if this assumption no longer holds, the central question is not which model is best, but which can be accessed at all. With frontier capabilities becoming an issue of national security and diplomacy, governments will be tempted to pursue "AI sovereignty" through the development of national champions or domestic alternatives to the leading US options (ChatGPT and Claude). But as understandable as this instinct is, it risks addressing the wrong problem. After all, AI is advancing too fast for such a strategy to pay off. Technological advantages that appear decisive can vanish within months. Today's breakthrough becomes tomorrow's baseline. Models that dominate headlines for a few months are quickly matched or surpassed by competitors. Even countries willing to invest tens of billions of dollars in developing models face daunting odds when competing directly against the world's largest technology firms. That is why the question is not whether a country can build a frontier model, but whether it can secure reliable access to frontier intelligence wherever it emerges. The recent Anthropic case illustrates the point. If access to a leading AI model can be restricted overnight, dependence on any single provider becomes a strategic risk. That does not mean every country should build its own frontier model. But it does mean that no country can assume uninterrupted access to someone else's system. For America's closest allies, the first priority should be to preserve access. Countries such as Japan and other G7 members share not only democratic values with the US, but also deep security interests. Supporting the technological resilience of allies ultimately strengthens America's own strategic position. Moreover, AI remains an immature and rapidly evolving technology whose future trajectory is uncertain. Continued collaboration between American model developers and allied countries that contribute technology, talent, infrastructure, and markets will be essential for expanding the ecosystem itself. AI should not become a technology that is hoarded; it should be one that is developed together. In the emerging AI economy, competitive advantage will increasingly come not from owning a single model, but from being able to evaluate, select, and orchestrate many models. Organizations that can seamlessly switch among competing systems will be more resilient than those that depend on a single provider and expose themselves to too many vulnerabilities -- from technical failures and commercial disputes to geopolitical pressure. Just as countries seek to diversify energy sources and semiconductor supply chains, they will also diversify their AI providers. Yet orchestration alone is not enough. Countries must also be able to evaluate AI systems independently. Governments need the ability to assess which models are genuinely useful compared to others, and what risks particular systems may pose. Such judgments cannot be delegated entirely to foreign companies or foreign governments. This is why recently established AI safety institutes and national cyber-security agencies are becoming increasingly important. Rather than merely trying to regulate AI, their role is to provide the independent technical expertise needed for informed national decision-making. Countries that cannot assess AI systems for themselves will inevitably find their choices shaped by others. Relying on multiple AI models rather than a single provider is its own form of deterrence, because it makes coercion less effective. If governments and organizations can seamlessly switch between multiple frontier models, restricting access to any one of them loses much of its strategic value. Model orchestration becomes both an economic capability and a geopolitical asset. But diversification is not enough. To achieve true AI sovereignty, countries must be able to combine national data with multiple frontier AI systems, generate knowledge about what works, and translate that knowledge into decision-making. This last step -- sovereign decision-making -- becomes difficult when a country depends on others to interpret its most important information. If the analysis of defense intelligence, economic data, public-health risks, or critical infrastructure ultimately relies on external actors, political autonomy becomes constrained. The implications extend well beyond AI policy. Once access to intelligence becomes a geopolitical issue, the challenge facing middle powers is clear. Rather than trying to replicate US or Chinese capabilities, these countries must ensure that they have room for strategic maneuver while remaining deeply integrated into the global economy. This is a familiar challenge: Throughout modern history, successful middle powers have prospered not by isolating themselves from larger powers, but by maintaining strong alliances while preserving the capacity for independent action. The emerging geopolitical competition will revolve around who designs, finances, operates, and optimizes the critical infrastructure of the AI age: data centers, power grids, communications networks, logistics systems, ports, and digital public infrastructure. AI capability will become embedded in the systems on which modern societies depend. In this environment, the meaning of "safe and trustworthy AI" must evolve, too. Today's debates about AI safety often focus on technical questions about model alignment, transparency, bias, misinformation, or harmful outputs. These are important concerns, but for governments, businesses, and the wider public, trust requires something broader. A trustworthy AI ecosystem is one in which users can rely on continued access, retain meaningful control over their data, and avoid becoming dependent on political decisions made elsewhere. It is an ecosystem in which access is not conditioned on political alignment, and where participation does not require the surrender of digital sovereignty. Trust lies not in any one model, but in the institutions, governance arrangements, and international relationships surrounding it. As AI becomes critical infrastructure, reliability, resilience, and political neutrality may matter as much as raw model performance. The most powerful model is not necessarily the most valuable if access to it can be withdrawn abruptly, or if dependence on it constrains strategic choices. AI sovereignty lies not in building a national version of ChatGPT, but in preserving freedom of action in a world where access to intelligence itself is contested. It is about optionality, not ownership. Ren Ito is co-founder and chairman of Sakana AI. This article was distributed by Project Syndicate.

US curbs on AI spur European firms to spread the risk

PARIS, June 22 (Reuters) - Limits on access to some U.S. artificial intelligence services are pushing major European companies to speed up spreading risk across multiple providers and reinforcing the need for more domestic alternatives. Recent restrictions include the U.S. government ordering San Francisco-based Anthropic, the company behind AI chatbot Claude, to suspend access to its Fable 5 and Mythos 5 models for foreign nationals, citing national security ?concerns. Such limits highlight a vulnerability for companies reliant on proprietary AI services delivered remotely - as those services can be restricted by their providers and cannot be run independently on a company's own servers. Executives from Siemens, Renault Group, Orange and ChapsVision told Reuters at last week's VivaTech conference in Paris that they already use a mix of U.S., Chinese and European models to avoid dependence on any one provider. Siemens said it uses Chinese models DeepSeek and Alibaba's Qwen, as well as Nvidia's Nemotron alongside other U.S. and European models. EU officials have been trying to reduce dependence on U.S. technology, which they see as a threat to the region's economic future, and have compiled a sovereignty package of measures to strengthen the bloc's capacity in semiconductors, AI and its digital autonomy. Yet, major companies say sovereignty is more about choice and diversity. "You need flexibility," Cedrik Neike, chief executive of Digital Industries at Siemens, told Reuters. "Sovereignty often gets confused with autarky (economic self-sufficiency), and autarky is absolutely not the way to do it." Europe's general-purpose AI providers, however, are few, headed by France's Mistral, while others like translation specialist DeepL have built strong positions in narrower fields. The artificial intelligence market is broadly split into two camps: open-source or open-weight models, which companies can run on their own servers, and proprietary models, which are accessed remotely and remain under the developer's control. "Today, in open source, when you look at European models, they're not impressive. At one point, the Americans were there, then they moved to closed source, and now there are only Chinese models in open source," OVHcloud Chief Executive Octave Klaba said. COST IS ANOTHER PRESSURE POINT Orange said its infrastructure could run all open-source models, including those developed in China, and described the risk in simple terms: running a Chinese model on European infrastructure is like buying a painting in China and bringing it home, because the model is self-operated and does not send data abroad when run on premises. The Anthropic restrictions, Orange said, made "patently clear, if it wasn't before, how important it is for Europe to have access to an AI service that it can control, that will never be switched off on a whim." Orange's Chief Executive Christel Heydemann, in a keynote address, called on Europe to build artificial intelligence it can access, govern and challenge on its own terms. French AI and data analytics company ChapsVision, which has won government contracts in France and Germany to replace U.S. rival Palantir, said it uses models ranging from Mistral and Anthropic to OpenAI and Qwen. For ChapsVision, sovereignty means having a credible alternative ready if a key service is cut. Software companies SAP and Sopra Steria agreed resilience comes through diversification, not isolation. IT group Capgemini said most AI providers were adapting their offerings beyond remote access to ease dependency concerns in Europe because the market was too large to lose, although it flagged it was a work in progress. Cost is also emerging as a pressure point. Token costs - the fees charged per unit of information processed - are rising as companies move to systems in which software agents carry out tasks automatically. Orange said executives would be "obsessed with cost per token" by the end of this year, citing Uber as an example of a company that had burned through its 2026 token budget in just four months. Carmaker Renault Group works with Google, Microsoft, Mistral, DeepSeek and Dataiku, using both open-weight and proprietary models, although it says it is not using DeepSeek in a meaningful way yet. "Renault Group already has an in-depth reflection on the cost of AI tokens, which have risen sharply and are pushing us to adapt," a spokesperson said. Rudy Kahn, a senior executive at German software firm Celonis, whose clients include BMW and Siemens, said companies need to first build the necessary infrastructure to give AI agents context on how their business works before deploying them. "If you do not provide a context model, AI needs to extract every single fact from the data itself," he said. "This will just blow your token bill completely." (Reporting by Leo Marchandon. Additional reporting by Dominique Patton and Gilles Guillaume; Editing by Matt Scuffham and Susan Fenton)

The Anthropic Case Changes Everything for Europe

For years, Europe believed that artificial intelligence was just another service. Then, in a matter of hours, Washington proved that an AI used worldwide could be cut off by a simple political decision. Just a few days ago, analysts in the cybersecurity centers of major corporations were opening their Anthropic interfaces as they did every morning, expecting to find the same tool they used the day before to track software vulnerabilities. But within hours, two of the most advanced models on the market, Claude Fable 5 and Claude Mythos 5, were withdrawn from global circulation. Not because they were broken. Not because they lacked server capacity. But because the U.S. government decided they should no longer be accessible to foreign nationals. Anthropic says it received the directive on June 12 and had to cut access for all its clients to comply with the order. For years, European digital sovereignty often seemed somewhat abstract, invoked in speeches about the cloud, semiconductors, or data. The Anthropic case changes the game. A cutting-edge technology, used by foreign companies and institutions, can be deactivated overnight by a political decision in Washington. Mythos 5 and Fable 5 were presented by Anthropic as a new generation of models, positioned above the Opus range in terms of capability, designed to solve the most demanding tasks in software engineering, analytical work, and especially cybersecurity. Anthropic explained that the Mythos-class opened a technological revolution, with defensive uses capable of identifying vulnerabilities in critical codebases on a massive scale, while acknowledging that this same power could become dangerous if it fell into the wrong hands. From the Chip War to the Model War To fully understand, one must look back a bit. Since 2022, the United States has already been using export controls as a central instrument of its technological rivalry, particularly toward China. Until now, the dominant logic was to control hardware, equipment, and physical flows. With Anthropic, Washington is changing both scale and nature. It is no longer just computing power that is becoming sensitive, but access to the model itself. Commerce Secretary Howard Lutnick invoked a risk that Mythos and Fable could be used by military intelligence services in China, Russia, or other countries deemed concerning. Anthropic went further by suspending the export of these models to all global destinations and to any foreign person, wherever they may be. This would be the first time the Department of Commerce has used the powers of the Export Control Reform Act of 2018 in this manner against an AI model. This shift in American doctrine has an immediate political effect. It shows that in the era of AI, the boundary between infrastructure, software, and foreign policy is becoming porous. Controlling chips already allowed for slowing down a rival. Controlling access to the model allows for deciding, in practice, who can benefit from the best layer of intelligence available. Anthropic, Caught in Its Own Risk Logic Yet, Anthropic is not the most reckless startup in the AI sector. On the contrary, its public communication has been built for years around caution, safety, and the control of the most dangerous capabilities. Its own Glasswing program, launched in April and expanded on June 2, was based precisely on the idea that a model as powerful as Mythos should first be reserved for a hand-picked circle of partners capable of using it defensively on critical infrastructure. Anthropic claimed that early partners had already identified more than 10,000 high or critical severity vulnerabilities, and that new participants came from more than 15 countries. A few days later, on June 9, the company nevertheless took a step forward. It publicly launched Fable 5, a "secured" version of Mythos, while continuing to offer Mythos 5 to certain Glasswing partners and preparing a broader trusted access program. Anthropic then described Fable 5 as a high-performance model, capable of working autonomously longer than its previous Claude models, with powerful uses in engineering, analysis, and vision. It also insisted that Fable and Mythos share the same base, with only the safeguards truly differentiating the two offerings. Three days later, everything stopped. Anthropic explained that the administration believed it had knowledge of a "jailbreak" method allowing the bypass of Fable 5's protections to identify software vulnerabilities. The company disputes the severity of the problem: according to Anthropic, the demonstration received only concerned a small number of already known and relatively minor flaws that other public models can also detect. It added that it had subjected Fable to thousands of hours of monitoring with internal teams, third parties, and the U.S. and British governments, without any "major risk" being found. Washington responded that in matters of national security, this was not enough: the government now reserved the right to purely and simply prohibit access. From a Technical Conflict to a Political Reckoning If the case has taken on such proportions, it is also because it fits into an already deteriorated relationship between Anthropic and the Trump administration. Earlier this year, the company claimed to have reached a breaking point with the Department of War regarding two uses it refused: mass surveillance of Americans and fully autonomous weapons. In a statement, Anthropic explained that the government then threatened to designate it as a "supply chain risk," a qualification it considered historically reserved for American adversaries and legally questionable for a domestic company. Relations between Anthropic and the administration were therefore somewhat strained earlier in the year after this refusal, and the government reportedly responded by placing the company on a national security blacklist. Added to this is another even more troubling factor: Amazon, despite being a major strategic partner of Anthropic and a massive investor in the company, saw its CEO Andy Jassy raise concerns with high-ranking administration officials regarding the security risks associated with the new models. The result is a case where it becomes difficult to completely separate the technical from the political. Officially, the heart of the problem is a possible security issue. In practice, several lines intersect: concerns over cyber-offense, the desire to reserve the best capabilities for an American perimeter, tensions over military uses, and the more general climate of an executive branch that treats AI as an attribute of power. This also explains the protest coming from within the American ecosystem itself. More than 80 leaders and cybersecurity experts signed an open letter calling for the lifting of restrictions, arguing that withdrawing the best available tools without a clear framework amounted to weakening security rather than strengthening it. Europe Discovers the Price of Dependency At the beginning of June, even before the Anthropic case, the European Commission proposed the Cloud and AI Development Act and a Chips Act 2.0 to support European alternatives in cloud, AI, and semiconductors. Ursula von der Leyen explained then that Europe could not afford to depend on others for the technologies that run its hospitals, energy networks, and critical services. But at the same time, Europe remains massively dependent on American infrastructure, whether it be the cloud, chips, or foundational models. Even Mistral, presented as the main European champion, operates in an ecosystem where computing, components, and part of the partnerships remain largely anchored to the United States. After the Anthropic Episode, Sovereignty It is possible that the ban will be partially lifted in the coming weeks. Discussions continue between Anthropic and the U.S. administration, Donald Trump told the G7 that negotiations were going "well," and the idea of a "trusted partners" status is already circulating to allow limited access to certain allies. It may even be, in the short term, that the case leads to a fairly classic technical and regulatory compromise: better security protocols, licensing procedures, or a finer gradation between authorized and prohibited uses. But even if access returns, the precedent will remain. European companies have learned that a cutting-edge American model can be interrupted for foreign users without a clear public procedure, based on a national assessment of the threat, even when those users are allies, economic partners, or cybersecurity actors. We may be entering a world where the most powerful AI models will be treated like advanced semiconductors, satellite systems, or certain defense capabilities: no longer as universally marketable goods, but as resources with conditional access, framed by nationality, alliance, political trust, and strategic compliance. If this is the case, then European digital sovereignty changes meaning. It is no longer just about "having our champions." It means, more bluntly, ensuring that the building blocks of intelligence indispensable to the economy, cybersecurity, health, or administration cannot be withdrawn from the outside.