The illustrations here are from me, but some of these were adopts. Character design credits:

* Goldkin (raptorized design): https://bsky.app/profile/tempelina.bsky.social
* Ant dragon adopt (fourth featured character): https://bsky.app/profile/razigator.bsky.social
* Pigeon dragon adopt (fifth featured character): https://bsky.app/profile/yotecoyote.bsky.social

PS: there's a lot of exonerative language in their AI policy stating that this isn't being used to train their model or third parties. It's useful to know that, assuming their policy is followed, that's likely true for now.

But it can be simultaneously true that this is aimed to (1) increase adoption of the feature without yet pulling training data and (2) that the policy can easily change later, under the guise of seemingly innocuous third-party changes.

Either way, worth opting out now if you're not okay with all of that.

And no, the setting is not present as best as I can tell anywhere else in their page, even where it shows the AI Companion feature as off. It's a special "opt out" hyperlink that appears at the top of the page, then disappears when clicked. I attempted to capture its contents but could not reproduce it after the fact, since I did not expect it to disappear and lack a second Pro account.

This way of doing things is absolute catnip for attackers, who serve fake links like this all the time via cross-site scripting (XSS). I took the extra step of verifying this wasn't such an attack, following the full attribution path for their email, and even with it all appearing correct and without surprising metadata that may indicate XSS, I'm still not convinced this is legitimate.

Anyway, absolutely shameful way of rolling this out. I get that there's no regulation for this right now and that this is probably Zoom engineers setting this up in the easiest way possible instead of deliberate nefariousness (gods, I hope), but please fucking do better than this.