AI: Could Germany adopt Anthropic?

The US government’s decision to banish AI giant Anthropic from its entire supply chain has handed Germany and Europe a “once-in-a-lifetime chance,” according to a politician from the center-left Social Democratic Party (SPD), junior partner in Chancellor Friedrich Merz‘s government coalition.

Anthropic, valued at $380 billion (€327 billion), has developed a series of large language models named Claude with a huge variety of applications, including piloting drones. The company has partnered with the US military and government since late 2024, and US media reports claim that Claude was used both in the raid on Venezuela and the bombing of Iran — though it was not clear exactly how.

In late February, US Defense Secretary Pete Hegseth demanded that Anthropic remove the clauses in its contracts that prohibit the use of Claude for mass surveillance and fully autonomous weapons.

When Anthropic co-founder and CEO Dario Amodei refused to comply, Hegseth declared the company a “supply chain risk” — a designation that had never before been applied to a US firm — and President Donald Trump ordered all government agencies to stop using Anthropic’s services. The company has now filed a challenge against the designation in court, saying that the blacklisting by the US government could cost it multiple billions of dollars as well as reputational harm.

There were two reasons why Anthropic would not accede to the Trump administration’s demands, the company said in a statement posted on its website in late February: “First, we do not believe that today’s frontier AI models are reliable enough to be used in fully autonomous weapons. … Second, we believe that mass domestic surveillance of Americans constitutes a violation of fundamental rights.”

A radical plan

Now Matthias Mieves, the SPD’s digital policy spokesperson, has written a letter to Chancellor Merz and European Commission President Ursula von der Leyen arguing that they should offer to bring Anthropic to Europe to develop its “human-centered, trustworthy AI” models. Anthropic’s refusal to comply with the US government’s demands showed that “the protection of the individual plays a role for Anthropic, and that fits the legal situation in Europe,” Mieves told DW.

Mieves’ letter, which he also posted on his social media profiles, went on to make three proposals for the German and European governments:

• Offer to make a major European city (Mieves suggested Berlin, Paris or Munich) a potential new base for Anthropic.
• Form an alliance of new European investors, including state bodies like European pensions funds and Euro-bonds, to ensure “digital sovereignty.”
• A guarantee from the European Commission and leading European nations to partner with Anthropic in future.

But experts in the AI field are skeptical that this is feasible. Daniel Abbou, head of Germany’s AI industry association BVKI, described Mieves plan’ as a “very, very nice idea,” but thought it unrealistic. “I personally would be extremely happy if Anthropic came to Europe,” he told DW. “But Anthropic’s main investors are, for example, Amazon and Google and it is deeply embedded in the US cloud infrastructure and capital markets. I’m not sure that such a move wouldn’t massively endanger the contracts and liquidity of Anthropic.”

Abbou also doubted whether losing the government contracts in the US really meant that all of Anthropic’s business there is about to dry up instantly. “Claude code and Opus 4.6 are going through the roof — they have huge access counts,” he said. “Of course, the pressure exerted by the US government in the military sector will have an effect on Anthropic, but it’s not like they can pack their cases and jump on a plane.”

What Europe needs in the AI sector

A much more viable plan, according to Abbou, would be to establish an “AI lab” in Europe in collaboration with Anthropic in order to create a kind of “reverse brain-drain initiative,” to stop Europeans from leaving to work at AI firms in the US.

What Europe really lacks, according to Abbou, is a better venture capitalism ecosystem. “There’s a lack of scaling possibilities in Europe that would attract venture capital,” he said. The biggest problem for the 600 German AI companies that the BVKI represents, he explained, was the financing of scaling-up operations. “That’s why so many of them go to the US, unfortunately,” he said.

Mieves is aware that his plan might be a little far-fetched — but, he said, it is supposed to sound radical. “Of course, it’s very difficult, and the probability that it will really work is low, but why did I make this proposal anyway? Because I believe we here in Europe need to think much bigger,” he told DW. “More than anything, we need to think in terms of radical measures if we don’t want to leave all the developments around AI to the US and China.”

The need for digital independence

The European Union is in the process of rolling out its Artificial Intelligence Act, part of its Apply AI Strategy, which is designed to foster “values-based regulation” of AI. This includes classifying new applications based on four levels of risk: unacceptable, high, limited, and minimal. But some companies are urging the EU to delay the law for fear that it will suppress early innovation and hinder the bloc’s self-professed hopes of “making the EU an AI continent.”

European politicians and businesspeople in the AI field routinely talk about the importance of Europe achieving “digital sovereignty,” meaning that data and expertise should be kept inside the EU, while safeguarding intellectual property.

Arthur Mensch, CEO of the French company Mistral, currently the only European company that can build large language models, said it was vital that Chinese and US companies be kept out of European industry. “[AI] is going to be running multiple digits of the GDP,” he told German public broadcaster DLF this week. “So we don’t want our industry to be running on American AI. I think that would be quite dangerous when it comes to the dependency that we are effectively creating. This means we need to be able to create models, it means we need to be able to operate the data centers in which they run.”

But even though everyone is aware of this problem, according to Mieves, politicians tend to “waffle on” about digital sovereignty without making plans that will strike that balance. “But there’s never any action, we never make a concrete plan, we never drive ambitious measures in order to actually become more sovereign,” he said. “That’s why I made my proposal, in order to trigger the discussion more forcefully about what we really need to do.”

Anthropic was repeatedly contacted for comment, but did not respond in time for publication of this article.

While you’re here: Every Tuesday, DW editors round up what is happening in German politics and society. You can sign up here for the weekly email newsletter, Berlin Briefing.