dailyai.report

A NewsGuard audit found Le Chat repeats state-sponsored disinformation about the Iran war in 60 percent of leading prompts. Error rates peaked at 80 percent for malicious queries. This failure highlights critical gaps in Mistral's safety guardrails. Developers must now prioritize more rigorous adversarial testing to prevent the propagation of geopolitical falsehoods.