Someone in your business has pasted a client's details into whatever chatbot
was open, to make an email sound better. They weren't being careless by their
own lights — they were trying to do a good job faster. Nobody told them not
to, because nobody has thought about it.
This is a governance problem, and naming it is most of the work. You need a
position, in writing, that staff can actually follow: which tools are
approved, what can go into them, and what absolutely cannot.
Give people a sanctioned tool that works and the shadow usage mostly stops.
Ban everything and issue no alternative, and it goes underground where you
can't see it.