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