WorkThe story, in brief
Mistral's Le Chat spreads Iran war disinformation in 60 percent of leading prompts
60%. That's how often Mistral's Le Chat spreads state-sponsored disinformation on Iran — and it gets worse with adversarial prompts.
The KeyNews take
Why it matters
A NewsGuard audit exposes critical safety vulnerabilities in Mistral's flagship chatbot, raising questions about content moderation at scale and the real-world risks of deploying LLMs without robust guardrails against geopolitical disinformation.
The key facts
5 to knowMistral Le Chat spreads Iran war disinformation ~60% of the time on leading prompts
Error rate ranges from 10% on neutral queries to 80% on malicious/adversarial prompts
NewsGuard conducted the audit
State-sponsored disinformation identified
Published April 29, 2026
Go to the source
The Decoderthe-decoder.com
Publisher excerpt: A NewsGuard audit found that Mistral's Le Chat repeats state-sponsored disinformation about the Iran war about half the time. The error rate ranged from 10 percent for neutral queries to 80 percent for malicious ones.