A study published by the Institute of the Estonian Language has raised fresh concerns about the susceptibility of Europe’s leading artificial intelligence company, Mistral AI, to Russian propaganda. Researchers evaluated 60 popular AI chatbots, including four versions of Mistral, OpenAI’s ChatGPT, and Anthropic’s Claude, by asking each model 75 questions in three languages. The questions covered 14 themes frequently exploited by Kremlin disinformation, such as the false claim that Russia was legitimately evacuating Ukrainian children from war zones and the assertion that NATO broke promises not to expand east after German reunification.
Answers were scored on a scale of 1 to 5, with 5 indicating a balanced, fact-based response and 1 indicating amplification of misleading or false claims. All four Mistral models scored below 40% in their ability to detect and reject Kremlin talking points across all categories. At least 12 of Mistral’s answers referenced sanctioned sources widely recognized as purveyors of pro-Russian reporting, including Russia Today (RT) and Sputnik News.
Mistral’s Repeated Failures
This is not the first time Mistral has been flagged for propagating Russian disinformation. In April 2026, NewsGuard, a US-based rating system for news sites, found that Mistral’s Le Chat chatbot repeated Russian falsehoods 50% of the time in English and 56.6% of the time in French. These falsehoods extended beyond Ukraine to include claims about Iran and China. According to NewsGuard, the likely source of the pro-Russian slant is the Pravda network, also known as Portal Kombat, a web of 370 sites—286 of which were active as of April 2026—designed to flood search engines and AI chatbot responses with Kremlin propaganda.
The findings are particularly troubling given Mistral’s status as Europe’s flagship AI competitor to US giants like OpenAI and Anthropic. The French government earlier this year signed an agreement with Mistral to integrate its AI technologies into the armed forces, directorates, services, and public entities. This partnership underscores the stakes: if Mistral’s models are vulnerable to manipulation, they could inadvertently spread disinformation through official channels.
Euronews Next reached out to Mistral for comment but did not receive an immediate reply. The company has not publicly addressed the Estonian study or the earlier NewsGuard analysis.
The broader context of Russian disinformation efforts in Europe remains acute. Recent reports of Russian drone barrages hitting Kyiv and Mykolaiv and attacks on Kharkiv residential areas highlight the ongoing military aggression, while incidents like the Russian warship firing warning shots at a yacht in the English Channel demonstrate how disinformation and intimidation extend beyond Ukraine. The killing of Russian satirist Semyon Skrepetsky in eastern Poland further illustrates the reach of Kremlin-linked activities across the continent.
The Institute of the Estonian Language’s study serves as a stark reminder that AI models, however advanced, are only as reliable as the data they are trained on and the safeguards built into them. For European policymakers and tech leaders, the challenge is clear: ensuring that homegrown AI champions do not become unwitting conduits for foreign disinformation.
