Europe’s most prominent artificial intelligence model is among the most vulnerable to Russian propaganda, according to a new study.
Researchers at the Institute of the Estonian Language tested 60 popular AI chatbots, including Mistral AI, OpenAI’s ChatGPT, and Anthropic’s Claude. The models answered questions across 14 themes, including false Kremlin narratives about Ukrainian children and NATO expansion.
All four Mistral versions scored below 40% in detecting Kremlin talking points. The study also found Mistral cited sanctioned, pro-Russian outlets like Russia Today and Sputnik News in at least 12 responses.
This is not the first failure. An April 2026 analysis by NewsGuard found Mistral’s Le Chat repeated Russian falsehoods half the time in English and nearly 57% of the time in French.
According to NewsGuard, the Pravda network-a sprawling web of hundreds of sites-likely floods AI tools with propaganda, influencing chatbot outputs. Mistral did not respond to requests for comment.
The findings carry geopolitical weight. The French government recently signed an agreement integrating Mistral’s AI into its armed forces and public entities. The company has also secured deals with Airbus, BMW, and the government of Luxembourg.