Anyone with half a brain should know by now (from experience if nothing else) that current AI is unreliable. At a foundational level, it is based on probability --- not logic.
Introducing additional probability into your workflow is literally rolling the dice. And probability itself shows that it is only a matter of time before it goes wrong.
In other words, decisions based on probability rather than logic are legally questionable.
https://tech.co/news/list-ai-failures-mistakes-errors
https://www.liveinsurancenews.com/california-health-insuranc...