AI tools can sometimes produce hallucinations—incorrect or misleading...

https://post-wiki.win/index.php/Why_Do_AI_Chatbots_Sound_So_Confident_When_They_Are_Wrong%3F

AI tools can sometimes produce hallucinations—incorrect or misleading information—with rates ranging from 22% to 94% depending on the context. For family offices, this risk is significant, especially when 20% of outputs may contain material errors