WWDC 2026: Apple Introduces Siri AI and the Next Generation of Apple Intelligence

At WWDC 2026, Apple introduced Siri AI as the centerpiece of the next generation of Apple Intelligence. The company presented it as a complete rebuild of Siri for iPhone, iPad, Mac, Apple Watch, and Apple Vision Pro, alongside iOS 27, iPadOS 27, macOS 27, watchOS 27, visionOS 27, and tvOS 27.

The word doing the work in that sentence is “rebuild.” Apple acquired Siri as a standalone app in 2010 and shipped it as a beta feature on the iPhone 4S in 2011, making it the first voice assistant most people ever used. It also spent much of the following decade as the assistant users complained about most, capable of timers and weather but unreliable at anything that required understanding context or chaining steps together. Any announcement about Siri is therefore measured against fifteen years of accumulated expectation rather than against the state of the art alone.

There is more recent history too. Apple demonstrated a version of exactly these capabilities — personal context, onscreen awareness, and in-app actions — when it announced Apple Intelligence at WWDC 2024, then acknowledged the following year that the personalized Siri features needed more time and pulled them from the near-term roadmap. That episode is why the 2026 announcement reads as a delivery rather than a debut, and why the demonstrations invited more scrutiny than a first showing normally would.

Apple says Siri AI can answer questions about what is on screen, search across messages, emails, photos, and other personal content using personal context understanding, and take more systemwide actions across apps. Apple also introduced a dedicated Siri app for revisiting conversations across devices, support for web-based answers using broad world knowledge, and tighter integration with Dynamic Island on iPhone and Spotlight on iPad and Mac.

Those three claims are technically distinct, and they fail in different ways. Answering about what is on screen requires a model to interpret interface content rather than plain text. Searching across messages, mail, and photos requires a semantic index of personal data that is built and kept current on the device itself. Taking systemwide actions requires apps to expose what they can do in a machine-readable form — the role App Intents has played in Apple’s developer frameworks — so the assistant can invoke a capability instead of simulating taps. The third is the hardest to make dependable, because a wrong answer is a nuisance while a wrong action changes something.

On the technical side, Apple ties Siri AI to the next generation of Apple Intelligence and its Apple Foundation Models. The company says the assistant runs on device when possible and uses Private Cloud Compute for larger requests, with the same privacy model Apple has been promoting across its AI stack.

That split is a genuine engineering constraint dressed as a privacy feature, and both readings are fair. Models small enough to run in a phone’s memory budget are less capable than models running in a data center, so any on-device assistant has to decide, request by request, what it can handle locally and what it must send away. Apple’s answer since introducing Private Cloud Compute in 2024 has been to route the larger requests to servers built on its own silicon, with the stated commitment that personal data is not retained there and that the server software can be inspected independently. It is a stronger claim than the industry norm, and it is also what makes the on-device tier worth optimizing so heavily.

Developer testing began on June 8, while a broader beta for users is scheduled for later in 2026. Apple also published a same-day update saying Siri AI would not ship in the European Union on iPhone and iPad with iOS 27 and iPadOS 27 because of DMA-related constraints, though Mac and Vision Pro support in the EU remains planned.

The shape of that EU exception is informative. The Digital Markets Act applies to designated core platform services rather than to a company as a whole, and iOS and iPadOS were designated while macOS was not — which lines up precisely with the platforms Apple says are and are not affected. Apple has taken this route before, delaying European availability of features while it worked through interoperability requirements, and has generally shipped them later rather than never. What that pattern does not tell European readers is how long later means in this case; Apple gave no date.

Two dates in the release are also worth separating. A developer seed in June is a first build handed to people writing software, not a preview of finished behaviour, and Apple has historically staged WWDC features across the releases that follow — some arriving with the initial version in the autumn, others in point updates, and occasionally one slipping past the cycle entirely. The public beta later in 2026 is the milestone that matters for most readers, because it is the first point at which the assistant meets ordinary usage rather than curated demonstrations.

That is ultimately the test. Assistants are not judged on the breadth of what they can do but on how often they do it correctly without supervision, and the features Apple described — reading the screen, searching a personal archive, acting inside apps — are exactly the ones where an error is immediately visible to the user. The announcement establishes what Apple intends Siri to be. Whether it is that thing is a question only the public beta can answer.