Restrictions imposed by the US government on access to certain artificial intelligence (AI) services are prompting major European companies to accelerate efforts to diversify their AI providers and increase reliance on domestic alternatives. These moves come amid rising concerns over dependence on proprietary AI models hosted and controlled by foreign providers, which can be restricted or suspended without prior notice.

The US recently ordered Anthropic, a San Francisco-based AI company known for its chatbot Claude, to suspend access to its Fable 5 and Mythos 5 models for foreign nationals, citing national security concerns. This decision has underscored the vulnerability of companies dependent on remotely delivered AI services that cannot be operated independently on their own servers.

At the recent VivaTech conference in Paris, executives from Siemens, Renault Group, Orange, and ChapsVision highlighted their use of a combination of US, Chinese, and European AI models to mitigate risks associated with overreliance on a single provider. Siemens, for example, uses Chinese-developed DeepSeek and Alibaba’s Qwen models alongside Nvidia’s Nemotron and other US and European models.

European Union officials have prioritized reducing dependence on US technology, viewing it as a strategic threat to the region’s economic future. To this end, the EU has introduced a sovereignty package aimed at strengthening domestic capabilities in semiconductors, AI, and digital autonomy. However, industry leaders stress that true sovereignty is about maintaining flexibility and choice rather than economic self-sufficiency. Cedrik Neike, CEO of Digital Industries at Siemens, emphasized that sovereignty should not be conflated with autarky.

The European AI market features a relatively small number of general-purpose providers, with France’s Mistral among the few leading firms. Others, like translation specialist DeepL, have established strong niches. The AI landscape divides broadly into open-source or open-weight models, which companies can host on their own infrastructure, and proprietary models offered through remote access under developers’ control.

Octave Klaba, CEO of OVHcloud, noted that European open-source AI models currently lag behind, with Chinese models dominating the open-source space after American firms shifted towards closed-source offerings. Orange, a major French telecommunications company, pointed out that running Chinese AI models on European infrastructure poses fewer data export risks compared to relying on foreign-hosted proprietary services. Orange’s CEO, Christel Heydemann, stressed the necessity for Europe to develop AI services which it can independently control and that cannot be arbitrarily switched off.

French AI and data analytics company ChapsVision, which has secured government contracts in France and Germany to replace US rival Palantir, said it employs a mix of models including Mistral, Anthropic, OpenAI, and Qwen. For ChapsVision, sovereignty means having reliable alternatives if key services become inaccessible.

Other companies, including SAP, Sopra Steria, and Capgemini, echoed the importance of resilience through diversification rather than isolation, noting that many AI providers are adapting their offerings beyond remote access to address European dependency concerns. Cost pressures are also intensifying, with firms monitoring the rising fees associated with high usage of AI services, commonly measured in tokens processed. Executives cited examples such as Uber, which rapidly exhausted its token budget for 2026.

Renault Group also combines proprietary and open-weight models from Google, Microsoft, Mistral, DeepSeek, and Dataiku, though it has yet to significantly deploy DeepSeek’s technology. Overall, these dynamics illustrate the growing urgency among European companies and policymakers to balance technological sovereignty, competitiveness, and flexibility amid a shifting global AI landscape.