From experience, people who tend to do this wouldn’t understand the issue even if they spent all the time in the world reading the email chain, or attending the meeting.

That’s what gives them a false sense that AI is helping, because AI is as good as themselves with comprehension and then saying plausible things that aren’t real. On the upside, it takes 2 seconds instead of an hour.

How far does it need to go to be a major incident?

I’m going to laugh in Java, where this has always been possible and reliable. Not like ai reliable, but expert reliable. Because of static types.