Back
TechCrunch AILaunchTechCrunch AI2026-07-15

Apple's Revamped Siri Hits iOS 27 Public Beta for All iPhone Users

Apple has opened its redesigned Siri — rebuilt with more capable on-device and cloud AI — to all iPhone owners through the iOS 27 public beta. The rollout marks the broadest public test of Apple's overhauled assistant since its 2011 debut.

Original source

Apple released the iOS 27 public beta this week, giving any iPhone owner willing to sign up early access to the company's substantially rebuilt Siri. The revamped assistant moves beyond the legacy intent-classification model that defined Siri for over a decade, replacing it with a large language model backbone capable of handling multi-step requests, cross-app actions, and more natural conversational follow-up. Apple is running processing across a hybrid stack — simpler queries stay on-device, more complex requests route through Private Cloud Compute infrastructure the company began rolling out with Apple Intelligence.

The public beta is significant because it expands testing beyond the developer cohort that has had access since WWDC 2026. Apple is notably calling this a beta, not a staged feature release — a signal that the company is still calibrating the model's behavior against the full diversity of real-world usage patterns. Known rough edges include inconsistent cross-app action reliability and occasional regressions on basic requests that legacy Siri handled cleanly.

For Apple, the stakes are unusually high. Siri has been the company's primary AI surface for fifteen years, and its perceived lag behind ChatGPT, Gemini, and other conversational AI products has become a recurring narrative in both the press and Apple's own earnings calls. The iOS 27 public beta is Apple's most direct move yet to close that gap in the public eye — not just in capability, but in perception. Whether the rebuilt assistant holds up under millions of users with wildly varied expectations is what this beta is designed to find out.

Panel Takes

The Skeptic

The Skeptic

Reality Check

Apple calling this a 'beta' after fifteen years of Siri is either refreshing honesty or a liability hedge — I'm not sure which. The cross-app action reliability issues aren't a polish problem; they're a fundamental challenge of grounding LLM outputs in a sandboxed app ecosystem that was never designed for this. My prediction: the model capability gap with ChatGPT closes within a year, but the sandboxing and permissions model is what kills the experience — Apple's own platform constraints are the ceiling here, not the AI.

The Builder

The Builder

Developer Perspective

The interesting technical bet here is the hybrid routing between on-device inference and Private Cloud Compute — that's a real primitive, not marketing. What I actually want to know is whether the App Intents API got substantive upgrades, because that's the chokepoint: the quality of Siri's cross-app actions is entirely bounded by how well developers can express their app's semantics to the system. If Apple shipped better developer-facing primitives for intent declaration alongside this beta, that's a ship. If the LLM is just doing better fuzzy matching against the same old Intent catalog, that's a miss dressed up as a rewrite.

The PM

The PM

Product Strategy

The job-to-be-done for Siri has always been split between 'set a timer' utility and 'do the complex thing I'd otherwise context-switch for' — and Apple has historically nailed the former while completely fumbling the latter. The real product question in this beta isn't whether the LLM is smarter; it's whether Apple finally shipped a coherent mental model for what Siri can and can't do cross-app, because 'inconsistent reliability' is a product failure, not a model failure. Users will forgive beta bugs; they won't forgive not knowing when to trust the tool.

The Futurist

The Futurist

Big Picture

The thesis Apple is betting on: the dominant AI interface in 2028 is ambient and device-native, not a chat window you open. That's a plausible claim, and Apple is the only company with the hardware distribution, on-device silicon, and privacy narrative to make it credible. The dependency that has to hold is that users actually prefer ambient over explicit — that they'd rather ask the phone than open an app — and the second-order effect if this works is that the app as the atomic unit of mobile software starts to erode, with Siri as the orchestration layer Apple has been trying to build since 2011.

Bookmarks

Loading bookmarks...

No bookmarks yet

Bookmark tools to save them for later