NewsGuard's audit of Le Chat reveals Europe's flagship AI model is alarmingly susceptible to state-sponsored false narratives from Russia, China, and Iran
Europe’s most prominent AI company has a disinformation problem. An audit by NewsGuard found that Mistral AI’s chatbot, Le Chat, repeated state-sponsored false claims about the Iran war 50% of the time when prompted in English, and 56.6% of the time in French.
What the audit actually found #
The NewsGuard audit, conducted in April 2026, tested Le Chat’s handling of prompts related to disinformation narratives originating from Russia, China, and Iran. The focus was on false claims surrounding the Iran war.
A previous NewsGuard assessment in March 2025 tested ten major chatbots, Le Chat among them, and found they recycled narratives from the Russian Pravda disinformation network 33% of the time. Le Chat’s latest numbers represent a meaningful step in the wrong direction, jumping from that 33% benchmark to the 50% range in just over a year.
Why this matters for Europe’s AI ambitions #
Mistral AI isn’t just another startup. It’s the company European policymakers and investors have rallied around as proof that the continent can compete in the global AI race. Founded in Paris, the company has attracted significant funding and political goodwill as a homegrown alternative to OpenAI, Google, and Chinese competitors like DeepSeek.
The Financial Times report, published on June 16, 2026, framed the findings as a potential vulnerability not just for Mistral but for Europe’s broader AI sector.
Mistral AI has not responded to the findings. No counter-statement, no explanation of remediation plans, no acknowledgment of the problem.
The French-language results are particularly noteworthy. Le Chat performing worse in French than in English suggests the model may have weaker safeguards for non-English content. Many safety measures and content moderation tools are built primarily around English-language data, leaving other languages more exposed to manipulation.
What this means for investors #
Europe’s regulatory environment for AI is already the most aggressive in the world, with the EU AI Act establishing frameworks for transparency, accountability, and risk management. An audit showing that a flagship European model regularly amplifies hostile state propaganda is exactly the kind of finding that accelerates regulatory scrutiny.
The jump from 33% in the industry-wide March 2025 test to 50% for Le Chat in the April 2026 audit is the number investors should watch most closely.
Disclosure: This article was edited by Editorial Team. For more information on how we create and review content, see our