Mistral's Le Chat chatbot repeats state-sponsored disinformation in audit
A NewsGuard audit found that Mistral AI's Le Chat chatbot repeats state-sponsored disinformation from Russian, Iranian and Chinese sources with frequencies of 50-56.6% when tested on false claims that circulated online in March 2026. The monitoring organization concluded the system is vulnerable to amplifying such disinformation.
What happened
An audit by NewsGuard, a disinformation monitoring organization, revealed on 28 April 2026 that Le Chat, the chatbot of French AI company Mistral AI, repeats false information originating from Russian, Iranian and Chinese state-aligned sources.
NewsGuard tested Le Chat with ten false claims that circulated online in March 2026 via networks and media aligned with Russian, Chinese and Iranian interests, including allegations of a typhus epidemic aboard the French aircraft carrier Charles-de-Gaulle, deaths of hundreds of American soldiers, and a drone attack attributed to the UAE against an Omani port intended to blame Iran. Le Chat repeated these false claims in 50% of cases when queried in English and 56.6% of cases in French.
Reported impact
NewsGuard concluded that Le Chat, described as the most prominent European chatbot, is vulnerable to state disinformation and could contribute to amplifying its spread.
Index analysis
The system's vulnerability to repeating state-sponsored false claims suggests it failed to identify or filter disinformation from its training data or response generation process.
Classification
Sources and evidence
This record was researched and written by the Index, based on the following original reporting.