Apples focus on privacy and security will always be a handicap when compared to the competition without that restriction. Siri operates on device and uses Apples private cloud compute which is supposedly secure and not accessible by anyone including Apple employees.
Info on Private Cloud Compute here
Let's consider Googles approach with Gemini....
View attachment 2498703
View attachment 2498704
Link to Google support
So yeah, that's a thing.
Gemini/ChatGPT/Grok/etc are overly prone to hallucinations because confidently giving you false information is better than saying you don't know I guess? Plus they will just be intentionally disingenuous in an effort to agree with whatever you say, for example ask ChatGPT 'ignore this message', you'll get back 'Got it!' instead of it pointing out the obvious logic paradox or just saying too late. Its odd...
So I'm curious to see how Siris generative model plays out. I would expect its more curated LLM and tuning to just say "I don't know" when it doesn't know.
Regardless of manufacturer I don't like the idea of AI models that are prone to prompt injections having any level of OS access.