Europe needs open-source to flourish
Europe needs to let its builders build. The bloc’s edge in AI isn’t making frontier models, it’s building on them, and open source is how.
The ever-elusive pursuit of creating European champions in the tech sector has become an obsession for political elites in the bloc, meanwhile the simplification measures the very same champions could benefit from are continuously diluted or delayed, as illustrated by the incomplete removal of overlaps in the AI Omnibus, or the pending data simplification package.
Meanwhile, builders keep on building. They often mention that while having European frontier AI models is desirable (France’s Mistral is a strong AI player, and it has contributed greatly to the open LLM stack), however, the Europeans can’t miss out on the opportunity to leverage and adapt the tools and services that already exist. That is especially important for thousands of smaller businesses running their own servers and models on-site and modifying them to their own standards.
What’s the EU’s open-source strategy about?
Europe’s advantage currently isn’t in building frontier LLM. But Europe has a major chance at building on the existing ones, and open-source is what allows entrepreneurs to flourish.
It has long been a strategy of EU institutions to support and implement open-source technology, as laid out in the EU Open Source Strategy, because it can easily reduce cost, avoid vendor lock-ins, and promote interoperability. But too often, rather than demonstrate how these solutions can serve as building blocks to greater innovation when mixed with European dynamism, EU officials have used open-source as an excuse to steer people away from working services, just because they don’t happen to be European.
This was made clear in a recent Politico story about Commission officials forced to use an open-source alternative to Microsoft Teams as their conferencing software, which employees panned and led to much derision online. This is the problem in a nutshell.
Using open-source tools should not be performative, serving to prove independence by using inadequate alternatives. Rather it should be about using proven solutions that have already been built and perfected elsewhere, and can serve as a building block to something greater. And making sure the biggest roadblocks for home-made solutions are cleared.
European entrepreneurs leveraging non-European models for European innovation
While major frontier AI models spring out of Silicon Valley and the labs and factories of China, Europe doesn’t need to reinvent the wheel when using LLMs or conferencing apps.
An interesting analysis put together by the Berlin-based online clothing retailer Zalando underscores these points. Businesses use a mixture of AI services to achieve what they need. Sometimes, they are local AI models hosted on their premises. Often, they are frontier models they can prompt from neoclouds. Entrepreneurs need flexibility, and open-source allows them this from the start.
Many German tech companies have taken this advice to heart, including firms like Deepset and Black Forest Labs, which build, modify, and launch open models for their AI products.
A July 2026 Bitkom survey of 102 German tech startups found 65% build their own applications on top of large AI models and 30% adapt existing models through training or fine-tuning. The vast majority of the innovation and value won’t be just in the models, but in the tech and tools themselves.
Even traditional companies like Bosch and BNP Paribas have taken to implementing open models and open-source technology in their businesses, customizing them to fit their needs and allow them to adapt to deliver value.
Products like NVIDIA’s Openshell provide secure sandboxes for autonomous agent testing, a technological solution to supposedly rogue AI agents escaping poorly determined sandboxes. That is open-source code built for everyone. SAP Germany, for example, has now begun using OpenShell in its own agent platform and is contributing code back to the project.
Using open-source tools, Europe can be a power player just as much as China and the U.S. And it’s not just about tech companies that are specifically working on AI, but those that implement it.
Published on EU Tech Loop (archive #1)