Exactly. And it's unreasonable to expect the public to be aware of the sheer amount of data that modern ML consumes, compared to how the more familiar "this call may be recorded for training purposes" is used.

Plus, the latter actually warns you about it every time you are recorded, not somewhere beforehand buried in a TOS. And the experience of "just talking to a computer" (not a person) makes people likely to share all sorts of private stuff.