Ecosia, the Berlin-based search engine that has built its brand on climate action and independence from non-European tech giants, has decided to stop using Mistral, the French company seen as Europe’s great hope in artificial intelligence. Its founder and chief executive, Christian Kroll, explained it in an interview with POLITICO published on October 6, 2026: “We are disappointed with the quality of Mistral.” According to Kroll, the French lab’s models are now “a year behind” the competition.
The decision carries symbolic weight. Ecosia is used by institutions such as Germany’s federal environment ministry and the UK’s National Health Service, and only in May had it replaced America’s OpenAI with Mistral as its artificial intelligence provider, precisely to strengthen its European profile. Five months later, the partnership is over and the search engine is turning to open-weight models, several of them developed in China.
Kroll listed several reasons. The first was technical: Ecosia suffered recurring problems, including overloaded servers. “We were simply too large a customer for Mistral,” he said. The second relates to its environmental goals, given the large share of nuclear power in France’s electricity mix. And the third concerns sovereignty: “The fact that Mistral relies on investment from international investors is, in our view, not truly sovereign,” Kroll said.
The new partner is Melious, a German platform that hosts a range of open-source and open-weight models, including China’s Qwen, from Alibaba; GLM, from Z.ai; and Kimi, from Moonshot AI. “We’ve roughly cut our costs in half while improving quality and performance,” Kroll said. The models run on European infrastructure powered by low-carbon energy, so data is processed in Europe even though the models were developed elsewhere.
The news sparked an intense debate, and Kroll sought to clarify it a few days later. In an exclusive interview with EU-Startups published on October 9, he insisted that the change is not a turn toward Chinese technology, but toward more affordable, flexible and energy-efficient open models. “We’ll probably be using Melious.ai in future. Unlike Mistral, this is a European company that offers high-quality, affordable AI inference powered by green electricity,” he explained.
Melious CEO and co-founder Simon Jakobs defended the approach in the same publication. “Europe needs its own AI capabilities,” he said. “At Melious, we run open-weight models on European infrastructure powered by clean energy, outside the reach of the US CLOUD Act. Whether a model was built in Europe, the US or China, it’s simply a set of weights on our servers. It can’t send data back to its makers, and customer data is never retained or used for training.”
Mistral did not respond to POLITICO’s request for comment, but its co-founder and chief scientist, Guillaume Lample, did react at a press conference. “I would encourage them to test our new model as soon as possible. We can even give them early access today if they want,” he said. Lample argued that building great models takes time, computing power and science, and noted that Mistral offers customers the weights of its models and the option of customizing them together.
The exchange coincided with the launch of Mistral Large 4, the company’s new flagship model, unveiled that same Tuesday. Jordan Bardella, president of France’s National Rally party, went so far as to say the model “puts France (and Europe) back in the race for artificial intelligence.” Mistral is also in a phase of strong expansion: in September it closed a €3 billion round led by Samsung Electronics at a valuation above €21 billion, according to EU-Startups.
The most sensitive point is the use of Chinese models. Amid US national security concerns, several Western research organizations have pointed to censorship problems on politically sensitive topics. POLITICO cited a NewsGuard investigation that found the five leading Chinese-backed models repeatedly failed to correct pro-Beijing falsehoods. Kroll argues that such biases can be addressed through technical measures, unlike shortcomings in quality.
As he told EU-Startups, Ecosia can switch to alternative models when dealing with sensitive subjects, such as questions about Taiwan, and is working to eliminate political biases from its AI-generated answers. Not all experts see it as that simple. Rasmus Rothe, executive director of the German AI Association, told POLITICO that overt censorship can largely be eliminated through targeted retraining, but that subtle bias is harder because it is embedded in the training data.
“With a few hundred example questions, you can scratch the surface, but you won’t reach the foundation,” Rothe warned. The debate also has an ironic side that has been pointed out on social media: Ecosia cites environmental reasons against French nuclear power, which is low-carbon, while using models developed in a country whose grid relies heavily on coal, even though those models are run in Europe on clean energy.
Beyond the controversy, the decision reflects an idea gaining ground in Europe. Kroll sees open-weight models as a huge opportunity for the continent: although Europe missed the first phase of the generative artificial intelligence boom, it can build competitive services on increasingly powerful open models without investing billions in training frontier systems. It is an approach that, according to POLITICO, EU leaders and advisers are also backing.
For Mistral, the episode is a blow to its image at a time when it is trying to establish itself as Europe’s technology champion. For Ecosia, which since 2018 has had an ownership structure that prevents it from being sold or distributing profits to shareholders, it is a bet on sovereignty understood as control over infrastructure and data, rather than the model’s origin. The open question is which of those two definitions of sovereignty will prevail in European artificial intelligence policy.
