European firms are increasingly turning to self-hosted open-source AI models as U.S. restrictions on proprietary services expose the risks of vendor dependence and strengthen the push for AI sovereignty.
European companies are accelerating adoption of open-source and open-weight AI models as recent U.S. restrictions on access to certain artificial intelligence services expose the risks of relying on proprietary platforms.
The shift gained urgency after the U.S. government ordered Anthropic to suspend access to its Fable 5 and Mythos 5 models for foreign nationals on national security grounds. The move underscored a key vulnerability of proprietary AI services: access can be restricted by providers or governments, leaving customers with limited control.
In response, major European enterprises are increasingly diversifying their AI portfolios. Siemens, Renault Group and ChapsVision told Reuters they already deploy a mix of U.S., Chinese and European models. Siemens uses DeepSeek, Alibaba’s Qwen and Nvidia Nemotron alongside other models, while Renault works with Google, Microsoft, Mistral, DeepSeek and Dataiku.
The trend is strengthening the appeal of open-source and open-weight AI, which can be deployed on local infrastructure and operated independently of external providers. Orange said its infrastructure can run all open-source models, including those developed in China, while stressing the importance of AI services that cannot be “switched off on a whim”.
The developments align with broader European efforts to strengthen digital sovereignty and reduce dependence on foreign technology. However, executives argue sovereignty should be achieved through diversification rather than isolation.
Despite the growing interest, Europe has relatively few general-purpose AI providers, with Mistral leading the field. Industry executives also noted the rising influence of Chinese open-source models such as DeepSeek and Qwen as enterprises seek greater resilience, control and flexibility.
