Apple’s approach to artificial intelligence has always been different — and iOS 27, previewed at WWDC 2026, makes that clearer than ever. Rather than chasing the largest models or the most internet-scale training data, Apple is betting on on-device intelligence, privacy-first architecture, and deep integration across its hardware ecosystem.
On-Device AI Gets a Major Upgrade
The headline feature in iOS 27 is the dramatically expanded on-device AI engine. Apple’s Neural Engine now handles models up to 7 billion parameters entirely on-device — triple the capacity of the previous generation. This means tasks like real-time photo analysis, live translation during FaceTime calls, and context-aware Siri suggestions happen without ever touching a cloud server.
During the WWDC keynote, Craig Federighi demonstrated the new “Memory Scout” feature: users can describe a vague memory — “that photo of my daughter at a birthday party with a blue cake” — and the on-device AI finds it in under two seconds by cross-referencing objects, colours, dates, and faces across the photo library. All processing stays local.
Siri Becomes Conversational
Siri has undergone its biggest transformation since the original launch in 2011. The assistant now supports multi-turn conversations with contextual memory. You can ask “What’s the weather in Amsterdam tomorrow?” followed by “Book me a lunch reservation near the office” and Siri understands “near the office” refers to your saved work address in Amsterdam.
More significantly, Siri can now take action across apps. A command like “Find the email Sarah sent about the Q3 budget, add the numbers to a new Numbers sheet, and share it with the finance team” works end-to-end — something no previous version of Siri could manage.
Privacy as a Differentiator
Apple is positioning privacy as its primary competitive advantage in AI. Every on-device AI feature ships with a privacy report accessible from Settings, showing exactly which data was accessed and confirming that nothing left the device. For cloud-dependent features, Apple’s Private Cloud Compute uses ephemeral, attestable servers that delete all data after processing.
This contrasts sharply with Google and OpenAI, both of which have faced regulatory scrutiny in the EU over data handling in their AI products. For European users — especially in privacy-conscious markets like the Netherlands and Germany — Apple’s approach could be decisive.
The Trade-off
Apple’s on-device-first strategy has a clear limitation: the most capable AI models simply cannot run on a phone. For creative tasks like generating images or writing long-form content, Apple quietly hands off to third-party models (currently ChatGPT) with user permission. The company has hinted at partnerships with additional AI providers, though no announcements were made at WWDC.
For the average user, the trade-off is probably worth it. Apple’s AI is fast, private, and deeply useful in everyday contexts. Just don’t expect it to write your next novel.
The Netherlands, for internationals
One email a week: the week's news for internationals in the Netherlands, new practical guides and what changed in the cities. No spam, unsubscribe any time.
