AI tools can sometimes produce “hallucinations”—inaccurate or misleading...

https://online-wiki.win/index.php/Should_a_Family_Office_Ban_ChatGPT_for_Client_Work%3F

AI tools can sometimes produce “hallucinations”—inaccurate or misleading outputs—with error rates ranging from 22% to as high as 94%. For family offices handling sensitive data, this risk is real