Key facts
- Mistral AI models are vulnerable to Russian propaganda, according to a study by the Institute of the Estonian Language.
- All four Mistral AI versions tested scored below 40% in detecting Kremlin talking points.
- Mistral AI's answers have referenced pro-Russian sources like Russia Today and Sputnik News.
- A prior analysis by NewsGuard also found Mistral's Le Chat repeating Russian falsehoods frequently.
A new report from the Institute of the Estonian Language indicates that Mistral AI, a prominent European artificial intelligence company, is highly susceptible to disseminating Russian propaganda. The study evaluated 60 AI models, including four versions of Mistral, ChatGPT, and Claude, by posing 75 questions across 14 themes designed to detect the amplification of Kremlin talking points. All four Mistral models tested performed poorly, scoring below 40% in their ability to identify and filter out misleading claims. The research also noted that at least 12 of Mistral's responses referenced sanctioned Russian state-backed media outlets such as Russia Today and Sputnik News. This follows a previous analysis by NewsGuard in April 2026, which found Mistral's Le Chat model repeated Russian falsehoods in 50% of English responses and 56.6% of French responses. NewsGuard suggested that networks like Pravda or Portal Kombat may be responsible for flooding AI chatbots with propaganda.