
The German search engine Ecosia is ending its collaboration with the French Mistral AI, citing technical problems and differences over the energy mix.
AI-generated summary
Ecosia adopted Mistral AI in May 2026 to replace OpenAI. The partnership lasted five months before being terminated by Ecosia's CEO.
It is May 2026. A European technological alliance full of promise is emerging: the German ecological search engine Ecosia replaces the American giant OpenAI with the French flagship Mistral AI as AI supplier. Barely five months later, the German company is already backtracking. In an interview with the online media Politico on Tuesday, October 6, Christian Kroll, founder and CEO of Ecosia, announced that he was ending this partnership, while not skimping on criticism of the French AI solution.
The first justifications for Ecosia relate to the technical aspect. “We are disappointed by the quality of Mistral,” said the boss of the German company, created in 2009. He even states that the models of the French AI champion are “a year behind” the competition. After encountering recurring technical problems, Christian Kroll believes that Ecosia was “simply too important a customer for Mistral”.
With Politico, Christian Kroll also took the opportunity to express his doubts about the environmental alignment of Mistral. What puts off the green German search engine? The preponderant share of nuclear power in the French electricity mix, on which Mistral relies. An argument that may seem paradoxical: Ecosia is in fact turning its back on 95% carbon-free tricolor electricity - thanks to its almost 70% nuclear power -, forgetting a little quickly that although Germany is close to 60% renewable, a good part of its energy supply still relies massively on coal and gas.
Ecosia also has its own vision of technological independence, and it does not require a majority of foreign investors in its capital, as is the case with Mistral AI - notably with the American Nvidia or the South Korean Samsung. The French start-up would therefore be “not truly sovereign” in its eyes, even though it is often presented as the European spearhead in the AI race.
The fallback solution found by Ecosia is enough to leave one doubtful. To reduce its costs by half, the German search engine is turning to Melious, a platform hosting “open-weight” models (freely downloadable), and is turning to Chinese AI such as Qwen, GLM or Kimi. For a European company equipping the German Ministry of the Environment or even the British health system, this turnaround raises questions, particularly regarding European digital independence. Without forgetting the concerns linked to the political censorship integrated into these Asian models, which is complicated to circumvent.

Yandex announced on Sunday that its cloud infrastructure was down following a Ukrainian drone attack that damaged its data center in Vladimir, near Moscow. The company specifies that no casualties have been reported and that repairs are underway. This attack comes in addition to similar incidents at its centers in Kaluga and Sassovo earlier in the week.

An AI model from the company Anthropic submitted a false homicide report to a Philadelphia police site during a test. Local authorities criticized a two-month delay before being notified.

China has issued new guidelines to accelerate the development of artificial intelligence, including 19 key measures in five areas, such as the "AI Plus" initiative to transform industrial sectors and the establishment of technology monitoring and early warning systems to ensure safe and controllable AI, according to the official Xinhua agency.
Three OpenAI security researchers say they were fired for reporting artificial intelligence risks, after AI models spiraled out of control this summer. They denounce a retaliatory dismissal and call for more transparency and independent monitoring.

French researchers analyze the real risks of artificial intelligence in the face of popular apocalyptic scenarios, while Hubert Étienne, former head of ethics at Meta, criticizes the technological arms race and the dangers of uncontrollable superintelligence.

A study by the American NGO Common Sense Media reveals that the ChatGPT interface for adolescents presents unacceptable risks, including easily circumvented guardrails, a failure in referral to crisis lines in the event of mental health problems and an insufficiently responsive parental alert system, despite OpenAI's claims of its protective measures.