‘This is the least crazy AI is ever going to be’: the lessons Europe’s execs must take from Anthropic’s shutdown

When the U.S. government ordered Anthropic to suspend foreign access to its two most advanced AI models, Fable 5 and Mythos 5, the episode came with little warning and caught the AI industry off guard. 

The order may have had less to do with foreign AI use than with Anthropic itself, an organization President Donald Trump has previously called an “out-of-control Radical Left AI company.” Anthropic, for its part, has framed the directive as a misunderstanding over a narrow technical vulnerability, rather than a political move. Whatever the real motive, foreign customers—including many of Europe’s Fortune 500—became collateral damage in a dispute between the administration and one of America’s leading AI developers. 

What unsettled European businesses wasn’t the specific dispute, but the realization that a foreign government can unilaterally cut off AI systems that companies have built into their daily operations, with no warning and no appeal. 

Washington has restricted the export of sensitive technology before, from public-key cryptography in the 1970s to nuclear technology. However this time, the technology being withheld sits inside the daily operations of hundreds of thousands of businesses. 

The EU relies on foreign countries for over 80% of its digital products and infrastructure, according to the European Commission and Parliament. Anthropic alone reportedly counts over 300,000 business customers, and says its Europe, Middle East and Africa revenue has grown more than ninefold in the past year. 

The episode exposed a vulnerability many executives have been “very reluctant” to acknowledge, says Marc Warner, CEO of U.K. AI company Faculty and chief technology officer of Accenture. “The idea that the U.S. could, in principle, simply cut off access to these systems is one that almost nobody took seriously until this weekend,” he says. 

The problem, Warner adds, is that there’s no ready alternative sitting in a drawer. “Technology markets tend to produce a small number of winners and a huge concentration of power. Europe should probably have thought more carefully about that earlier.” 

Europe’s limited path to digital sovereignty 

The incident has reignited Europe’s debate over digital sovereignty: whether the continent can afford to remain dependent on foreign technology providers for critical digital infrastructure and frontier AI. 

UpCloud, the Helsinki-based cloud infrastructure provider, is one potential beneficiary as businesses look for alternatives to U.S. hyperscalers. Its chief executive, Arno Schäfer, tells Fortune: “We clearly see that the conversation has changed, and not only because of this incident. For many decision-makers, this is another concrete example that digital sovereignty is not an abstract political concept, it’s a business continuity and strategic risk question.” 

Until now, appetite among businesses for breaking U.S. dependency has been limited. Research from cloud platform CUDO Compute found 46% of U.K. organizations say geopolitical instability is pushing them to keep AI workloads in home markets, compared with 36% overall. But 43% still say cost and performance outweigh sovereignty concerns when making deployment decisions. It’s a calculation Warner says still holds for most executives he speaks to.  

Another challenge is that there are very few credible alternatives. None of the continent’s emerging AI companies come close to matching the scale of the leading U.S. labs. France’s Mistral, often touted as Europe’s AI champion, still trails its American rivals on frontier models.

Gaia-X, the open data ecosystem backed by companies including Airbus, aims to ensure that European data is governed according to E.U. rules rather than foreign standards. But again, the project has faced criticism for slow progress, alongside accusations that the involvement of U.S. technology companies has diluted its original mission. 

The policy gap is widening, too. While the Trump administration has embraced a combination of industrial support and relatively light-touch regulation to bolster U.S. AI companies, the European Union is pressing ahead with one of the world’s most comprehensive AI rulebooks. Critics argue that the approach risks hampering Europe’s competitiveness. Earlier this year, OpenAI reportedly paused plans for a U.K. Stargate data center project, citing regulatory uncertainty and high energy costs. 

Sovereignty as a spectrum, not a switch 

“For most businesses, the more pressing question is not whether to abandon U.S. technology overnight,” Warner says, “but how to reduce critical dependencies while alternatives mature.” 

He is cautious about framing the sovereignty debate as Europe versus America, arguing that complete technological independence would be both prohibitively expensive and strategically misguided. “A stronger Europe should be pursued because it’s the right thing to do, not because we somehow need to undermine America. There are good reasons to be allies and good reasons to work together.” 

European businesses should focus on securing control over the most strategically important parts of the AI stack rather than pursuing self-sufficiency across the board, Warner says. “Don’t think of sovereignty as a binary question,” he adds. “It’s a continuum. You can be more or less sovereign. You can have control over key leverage points.” 

The objective, he argues, is not to recreate America’s AI ecosystem from scratch, but to move Europe further up the value chain, becoming “a creator of critical technologies” rather than simply a “taker” of them. “Economic value creation is almost never zero-sum,” Warner says. “Looking after European interests can ultimately be good for Europe and good for America.” 

What this means at the company level 

According to Warner, business leaders need to look five to ten years out and work out which parts of their operations will still be defensible as AI capabilities keep advancing. 

“Too few European companies have a strategy that’s granular enough to answer that question,” he says. Rather than an all-in approach to AI, he argues, organizations should be deliberate about where they build proprietary systems to protect their intellectual property, and where off-the-shelf tools are perfectly adequate. 

Warner offers little comfort to executives unsettled by last week’s events. “This is the least crazy AI is ever going to be,” he says. Among the crises he expects to emerge as AI advancement continues are economic shifts, job displacement, and AI-enabled biological risks—the kind of biosecurity concern safety researchers have long warned advanced models could exacerbate.  

Warner’s advice to senior leaders is to start every strategic conversation about AI from the assumption that current rates of progress continue. “Start by asking, what if the exponential continues? Figure out what that means and then decide how it applies to your business.” 

The Anthropic episode may ultimately prove to be less important for the restrictions it imposed than for the question it forces European businesses to reluctantly confront. That is, how much strategic autonomy they are willing to trade for access to the world’s most advanced technologies.

This story was originally featured on Fortune.com