Restrictions on access to certain American artificial intelligence services are prompting large European companies to move faster toward using multiple AI providers — and are strengthening the argument for homegrown European alternatives.
Among the most notable recent restrictions: the U.S. government directed San Francisco-based Anthropic, the company behind the AI chatbot Claude, to cut off foreign nationals from its Fable 5 and Mythos 5 models, citing national security concerns.
Those kinds of limits expose a significant weakness for businesses that rely on AI services delivered remotely. Unlike software run on a company’s own servers, these proprietary services can be shut off or restricted at any time by the companies that own them.
Executives from Siemens, Renault Group, Orange, and ChapsVision spoke with Reuters at last week’s VivaTech conference in Paris, and all said their companies already draw from a mix of American, Chinese, and European AI models to avoid being locked into a single provider.
Siemens, for instance, uses Chinese models including DeepSeek and Alibaba’s Qwen, along with Nvidia’s Nemotron and various other U.S. and European models.
European Union officials have been working to reduce the region’s reliance on U.S. technology, viewing that dependence as a risk to Europe’s economic future. They have put together a sovereignty package aimed at strengthening the bloc’s capabilities in semiconductors, artificial intelligence, and digital independence.
But major corporations say sovereignty is really about having options, not about going it alone.
“You need flexibility,” said Cedrik Neike, chief executive of Digital Industries at Siemens. “Sovereignty often gets confused with autarky (economic self-sufficiency), and autarky is absolutely not the way to do it.”
Europe’s lineup of general-purpose AI providers remains thin. France’s Mistral leads the pack, while others like translation specialist DeepL have carved out strong but narrower roles.
The broader AI market is divided into two categories: open-source or open-weight models that companies can host on their own infrastructure, and proprietary models that are accessed remotely and stay 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,” said OVHcloud Chief Executive Octave Klaba.
Orange said its infrastructure is capable of running all open-source models, including those from China, and put the risk in straightforward terms: using a Chinese model on European servers is similar to buying a painting in China and bringing it home — the model operates independently and doesn’t send data back to China when run locally.
The Anthropic restrictions, Orange said, made it “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, speaking at a keynote address, urged Europe to develop artificial intelligence that the continent can access, govern, and challenge on its own terms.
French AI and data analytics firm ChapsVision, which has secured government contracts in France and Germany to replace U.S. competitor Palantir, said it draws on models from Mistral, Anthropic, OpenAI, and Qwen. For ChapsVision, sovereignty means always having a reliable backup if a critical service goes dark.
Software companies SAP and Sopra Steria also agreed that resilience comes through diversification rather than isolation. IT group Capgemini noted that most AI providers are expanding their offerings beyond remote-only access to ease dependency worries in Europe, recognizing the market is too valuable to walk away from — though it acknowledged the transition is still ongoing.
Cost is increasingly becoming another pressure point for companies.
Token costs — the fees businesses pay per unit of information processed by an AI system — are climbing as more companies shift to automated AI agent systems that perform tasks on their own.
Orange said its executives would be “obsessed with cost per token” before the year is out, pointing to Uber as a company that burned through its entire 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, though it noted it is not yet using DeepSeek in any significant way.
“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 for the company said.
Rudy Kahn, a senior executive at German software firm Celonis — whose clients include BMW and Siemens — said companies must first build the right infrastructure to give AI agents context about how their business operates before putting those agents to work.
“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.”
