Andrew Jack / Financial Times:
Study: Mistral and other open-source AI models are among the weakest at filtering Russian disinformation; Mistral's top model ranks 47 out of 60 tested models — Open-source generative models are worse at removing false news than others, according to Estonian researchers
