Craig Federighi had a line in the keynote that was doing a lot of work: “Other tech companies are racing forward, pursuing AI for the sake of AI, without clear regard for the people.” Whether you read that as a genuine philosophy or a positioning move against Google and Microsoft, it set the frame for everything that followed. This was a WWDC about what kind of AI company Apple wants to be.
Here’s what I took away.
The biggest story is that Siri AI is finally real. The underlying model is a combination of Apple’s own foundation models and Google’s Gemini, deeply integrated across iOS, macOS, and the rest of the platform stack. It can analyze text, handle speech-to-text, generate images, and execute multi-step tasks across apps, all running on local hardware, with the heavier lifting handled by Private Cloud Compute when needed. Xcode got updates too, with Gemini now sitting alongside Claude and OpenAI as available options for developers.
Like what you’re reading? Become a free member →
Watching the demos, it’s hard not to notice that Siri is doing things Google Assistant has been capable of for years. I don’t think that’s entirely a knock on Apple; it’s more a knock on how innovation is stifled by corporate silos and private property. But Apple was behind, and the keynote made that obvious even as it was trying to celebrate the catch-up. What’s genuinely interesting is the local, on-device angle: a model that understands the full context of what you’re working on, without that context leaving your device. That’s a real differentiator.
Spatial Reframing is going to be more useful than it sounds. It’s an AI feature coming to the photo library that lets you adjust the angle the photo was effectively taken from after the fact. For anyone who has ever taken a shot where the composition was almost right, this is going to be huge. The kind of thing that sounds like a gimmick until it saves a photo you actually care about.
There was some speculation before the keynote about whether they would introduce the ability to describe a shortcut or a Safari extension and have AI build it for you. Turns out: yes. You can describe what you want, and the system handles the construction. Shortcuts has always had a steep learning curve, and that’s kept a lot of people from ever going deep with it. This doesn’t replace anyone who’s already doing complex automation, but it opens the door for many people who never walked through it in the first place. That’s a bigger deal than it might seem.
The lock animation got so much screen time during the keynote that I said you should take a shot every time it appeared. Apple was explicit: data processed through their servers is never stored, remains inaccessible to Apple or anyone else, and is used only to execute the specific request. They also announced that outside security researchers can audit the system at any time. That last part is significant. It’s one thing to make a privacy claim. It’s another to invite scrutiny. The reason this matters beyond Apple is that it addresses a real problem: many people haven’t fully leaned into AI tools because they genuinely don’t know where their data ends up. That’s a legitimate concern, not technophobia, and Apple is at least trying to answer it.
Siri AI is not launching on iPhone or iPad in the European Union or in China. Apple’s position is that the EU’s Digital Markets Act would require giving any AI assistant near-total device access, creating privacy risks Apple isn’t willing to accept. Federighi said Apple was “deeply disappointed.” There’s no timeline for resolution.
Here’s what’s true: both the EU and China have regulatory environments that complicate the deployment of privacy-first AI. Here’s what’s also true: the DMA exists because the US has a monopoly on technology, and the EU’s answer was regulation designed to give European capital a “fairer” shot. In the end, the real losers are EU iPhone users, who are waiting on a geopolitical stalemate with no clear end.
A few things worth mentioning that didn’t make the main list. The parental controls overhaul is significant and worth careful consideration. Features like these tend to assume that parents always act in their children’s best interests. That assumption doesn’t always hold. And the longer arc here is that Siri AI is pointed directly at the things most people use tools like Claude or other AI assistants for: analyzing files, organizing information, and helping you write. Apple missed the AI wave, and that’s no longer debatable. But they have something that no AI company has: the hardware where your actual life lives. If they can build a model that understands your context without compromising your privacy, that’s a different kind of play than any AI company is running.
Developer betas are out. Public betas come in July. Full release this fall. We’ll see what holds up.