It's a strange mixture, and it's anticipating future needs as well as current things like malware scanning. Probably Zoom and all the other mainstream meeting services will have to use AIs on audio and video streams in due course, in an arms race against other AIs being used to generate audio and video spam, fake job interview candidates, scam calls from people that look and sound like your own family, and such things. On the other hand, who wants their daily standups scrutinised by facial sentiment analysis and posted to monthly management reports, which is the sort of thing it seems Microsoft might do.

I don't think Zoom is at fault for wanting or needing to use AIs or plan to, on the streams. They might even be ahead of the game. But their communication about it is rather weasely at the moment, and their COO and CISO comments on here don't help with that.