As ever, disregard what a company says publicly and focus on what the legal documents says you are agreeing to when you use their product.
https://www.theverge.com/2023/8/7/23822907/zoom-train-ai-models-user-data-customer-consent
The Verge
Zoom says its new AI tools aren’t stealing ownership of your contentZoom says in a blog that it doesn’t use content from video calls on its platform “without user consent” following revelations that its TOS had been updated to seemingly give it broad user data control for AI work.
@carolannie @scalzi Basically this. When they can change the TOS essentially at will then they can do whatever under the cover of “using our services constitutes agreement”. They will do it without asking and only acknowledge it if they get caught…which will be difficult as the models mulch the data pretty effectively and the models are basically black holes.
A good rule of thumb: When a tech company promises (without legal backing) that they're not going to do "a thing," 75% of the time they're doing something extremely similar and weasel wording out of it, and 24% of the time they're just straight up lying because they're not getting legally called out for it
Even based on what they said in that article, they're not denying that they are using the content to train internal models. They're just not "claiming ownership." I'm sure that will be a huge relief to all the patients and healthcare providers (and HIPAA auditors) using Zoom.
Because as we know, training content for LLMs could NEVER leak across account boundaries...
Note the words 'without consumer consent'. These people have shown themselves to have such a loose and whimsical notion of consent that they would claim you've signed away your firstborn child by merely glancing at the first paragraph of an EULA. They've taken the fair folk in mind, and said, "we can do better than that."