Siri's long-awaited AI update has finally arrived

Apple's Siri Reboot: Useful at Last, or Just Catching Up?

Apple announced Siri AI at WWDC this week, and the short version is: it can now do things that ChatGPT has been doing for two years. That is not a criticism exactly, just context worth keeping in mind.

The update is unequivocally a big one by Siri's own historical standards. The new version taps into your emails, texts, photos and browsing history, pulls live data from the web, and surfaces it all through a standalone app or directly in the OS. So if you want Siri to find an address buried in a three-week-old message thread, or move a batch of photos into an album, it can now actually do that. These are things that felt embarrassingly out of reach before.

What Apple is going for here is AI that disappears into the operating system rather than demanding you open a chat interface. Francisco Jeronimo at IDC put it well:

"If Apple makes AI feel natural, private and useful for mainstream users, it will not just strengthen its ecosystem. It could redefine what consumers expect from every device they use."

That is the forward-looking version. The grounded one is that Apple is running this on Google's Gemini model under a multi-year deal announced in January. After years of presumably trying to build something in-house, they went and licensed it. Investors knocked the stock down close to 2% on the day, and that reaction derives less from the features themselves and more from a credibility gap Apple has built up through repeated delays and underwhelming AI rollouts.

Privacy as a Strategy, Not Just a Value

Craig Federighi's line about competitors "pursuing AI for the sake of AI" was pointed. Apple is positioning privacy as the differentiator: your data stays grounded in your device context, and cloud access is selective rather than wholesale. Whether that architectural choice actually limits capability compared to cloud-native AI is something worth watching once real users get hold of it. Apple says the US English will be released first, with EU and China launches delayed due to regulatory complications in both markets. That is a significant slice of their customer base sitting this out initially.

The Execution Problem

Daniel Newman at Futurum called this a "prove-it moment," and I think that framing is exactly right. The demonstrations at WWDC were videotaped and controlled, which tells you nothing about how Siri AI performs when you actually need it. Apple's track record on AI feature delivery has been inconsistent enough that the product has to earn back trust through actual use, not through keynote video.

The Gemini arrangement makes sense from a pragmatic standpoint. Building frontier AI models is expensive and slow, and Apple's core competency is hardware and ecosystem integration, not LLM research.

You can like to leverage Google's model while controlling the privacy layer and the user experience. The open question is whether a licensed model with Apple's constraints on top can keep pace as the underlying AI landscape keeps moving.