My reading of the fine print (IAAL, FWIW) is that turning off Gemini Apps Activity does not affect whether human review is possible. It just means that your prompts won't be saved beyond 72 hours, unless they are reviewed by humans, in which case they can live on indefinitely in a location separate from your account.

I also asked Gemini (not Ultra) and it told me that there is no way to prevent human review.

You should never ask an LLM to answer questions about itself. The answer is guaranteed to be hallucinated unless Google specifically finetuned it on an answer of that question. The answer it gave you is meaningless. (But also, coincidentally, correct.)

I recall seeing that OpenAI finetuned ChatGPT on facts related to itself, and I figured Google likely did the same. But you're right about not relying on its representations. I only skimmed its answer to see if it seemed consistent with my reading of the fine print.

You thought they would feed internal user data handling policy to a public facing LLM?

Yes, I thought they would feed the LLM's FAQ to the LLM. As I said above, OpenAI did this with ChatGPT. They even fed data about how ChatGPT was created to ChatGPT.

There's a thing that says even with activity off, they retain for 72 hours for "safety and reliability"

Seems like what any reasonably sized corporation would do with an entirely new product, based on entirely new and very unreliable tech.

Can you even access the opt out though before it hoovers up your messages and logs from the last 90 days? I didn't proceed because that didn't seem possible.

The word "future" in what you quoted is doing a lot of heavy lifting.

My assumption is you have to give it access to all your data, which it then processes for before you can opt out of it.

Why do you say that? I’ve never had that happen with any other of my Google data opt-outs.

I hate Google as much as the next person but, yeah, messing with opt-outs is something I've seen with Microsoft and Meta but not with Google.

That wouldn't be legal I think.

You'd think so, but these companies skirt around it by then adding or breaking up permissions even further, like "oh, yes you DID disable data collection for X, but.... we added a new permission for data collection for Y, and by the way it is opt-out! Too bad!".

LinkedIn is the master of this. They keep creating new notification types, which are enabled by default.

That's not how it works

They usually go for the "Software Bug nothing we could do." Microsoft and Meta are notorious for playing the system like that, with no recourse.

Usually? Can you share one example?

Search for Microsoft and settings reset in the search bar and you get ample examples.