Surprised no one has talked about the apparent hypocrisy of Apple’s privacy stance. Apple won’t sell your data for AI, but they will use AI models built from mined user data developed by other companies.
Got a tip for us?
Let us know
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
Apple's Google Gemini Deal Could Be Worth $5 Billion
- Thread starter MacRumors
- Start date
- Sort by reaction score
The non-“chat friend” parts of it seem to be quite capable in recent releases. That will feed into the image creation clip art for the apps plus creative video as well.
If Mark Gurman’s reporting is to be believed, Apple had a lot of this internally working last year at least 70% of the time using their own models. I would imagine they can close the gap once they swap in the Google models.
Yes financially it is, but it's not as if Apple is poor.
And this shouldn't just be about money (but likely is with Tim and shareholders)...it's about Apple innovation, and their ability to fully determine the future of Apple Intelligence.
The best solution for Apple. A lot of people me included called for this exact scenario 2 years ago. Apple does not need a consumer facing LLM or chatbot. Anyone can access their favorite version at a touch of a button and why would they want burn through $100's of billions on something that 1- no one is clamoring for, 2- Are not ready for Prime time, and 3- sucks $100's of billions of dollars from companies without generating anything in return. There is no ROI and the technology is not fully baked yet.
They’ll never catch up, like they’ll never catch up with the sales of Android devices. They’ll still make a decent chunk of change, though.
I've enjoyed the ChatGPT integration, though I'd love it if I had the option to have the ChatGPT app keep track of what I ask Siri. I'm a bit disappointed by a move to make Gemini fully integrated... if I could pick one, Claude is usually best suited for my needs (which of course will be different for everyone). Gemini has specifically been less useful/helpful than Claude or ChatGPT (though I am intrigued by the Notebook fork of Gemini).
Bigger question for end users is what version of Gemini is getting integrated into Siri and how far back will it be from the generally available Gemini? Will we be seeing a new version of the delayed Android release where Pixel phones are on version X and everyone else is on version X-1 for a year?
It's the non-friendly bits where I find it falls flat most often. As an example, it couldn't figure out an error I was getting when integrating fpdf2 - something Claude had no problem with.
Google can get a lot of value without Apple giving away anything in terms of user privacy.
One example would be statistics on the types of queries being made. Another would be the accuracy rating of queries and whether users are satisfied with the results they got (quality control).
Even statistics on processor/energy used vs number of queries to see how “efficient” the models are.
The idea is I don't want a single penny of mine going toward any google products or services, either directly or indirectly. My understanding was all iCloud data was stored on Apple hosted infrastructure. If it is in fact stored on Google infrastructure, I find that deeply disturbing.
I think a lot of people are blowing this out of proportion for nothing
Apple will likely use a custom, isolated version of Google’s Gemini that does not store, log, or train on Apple user data. Apple would treat Gemini as stateless compute, sending only anonymized, minimal requests with no Apple ID or identity attached. Unlike Microsoft, which retains user data to improve Copilot (while sandboxing it from OpenAI), Apple’s model prioritizes privacy over learning by ensuring neither Google nor Apple builds long-term memory from user interactions.
Apple will likely use a custom, isolated version of Google’s Gemini that does not store, log, or train on Apple user data. Apple would treat Gemini as stateless compute, sending only anonymized, minimal requests with no Apple ID or identity attached. Unlike Microsoft, which retains user data to improve Copilot (while sandboxing it from OpenAI), Apple’s model prioritizes privacy over learning by ensuring neither Google nor Apple builds long-term memory from user interactions.
Last edited:
What world are you living in where Meta doesn't exist?
Thankfully I have zero interaction with Meta and is easily avoided. Google is harder to avoid, especially if you email someone with a gmail address.
I was ready to buy an Occulus Rift until Carmack sold the IP to Zuck.
As long as I can turn it off I am okay with this. And hopefully the setting to disable is not buried like it is on my Pixel 8 Pro.
Presumably it's "low" because Google's not supplying compute. Apply is still going to supply the compute, Google is just supplying the model.
Google is such a huge part of tech/the internet/online advertising that not wanting a single penny to go to Google even indirectly seems like an impossible wish.
Have you tried it lately? It's gotten a lot better, fairly recently. I would expect the different services to leapfrog each other for a while.
It seems pretty likely Apple won’t hand user data over to Google, in the same way they protect user privacy in their arrangement with OpenAI. Honestly if Apple thinks Google provides the best backbone for their utilization of LLMs, I’m OK with it, so long as they go about it reasonably. It’s all back-end stuff that shouldn’t be particularly visible to users beyond the Siri/OS/application front-ends.
Almost like not buying a Nestle product. They have their fingers in everybody’s pie.
I imagine a lot of companies you do business with as a consumer pay Google for a variety of services. 🤷♂️
Yeah, I watched Gemini 3 Pro fail to implement some fpdf2 functions in a python thing I'm working on.
My workplace uses Gemini, so I toy around with it semi-frequently and I never really come away impressed.
