Apple shipped a working version of Siri AI in the iOS 27 consumer beta in July, delivering on features first promised more than a year ago and now built on Apple Foundation Models trained with help from Google's Gemini. The assistant understands personal context on the device, answers open-ended questions, and holds a natural voice conversation with adjustable pacing and expressivity. The public release is expected alongside iOS 27 in September.
The core upgrade is context. Siri can now pull up a specific receipt saved on the iPhone without being told where it lives, extract a driver's license number from a photo of the license, or resurface a QR code screenshotted weeks ago as a reminder. It reaches across photos, emails, contacts, texts, and calendar entries at once, which is what Apple originally pitched when it first previewed the personal-context feature.
“The AI assistant now does what Apple promised: It understands your personal context, taps into broad world knowledge to answer your questions, and surfaces relevant information stored on your iPhone.”— Sarah Perez, TechCrunch Consumer News Editor
Voice interaction has been rebuilt. Users can carry on back-and-forth exchanges with Siri, tune how expressive and how fast the voice responses are, or switch to typing. For the first time, Siri has a dedicated app that stores and references past conversations, closing one of the most-cited gaps between Siri and the ChatGPT and Gemini apps.
Key facts
- 01Siri AI shipped in the iOS 27 consumer beta in July, with the public release expected in September.
- 02The revamped Siri handles personal context queries across photos, emails, contacts, texts, and calendar without the user specifying where the data lives.
- 03Apple partnered with Google to use Gemini to train and refine Apple Foundation Models that run on Apple silicon and Apple's private cloud compute.
- 04Siri now supports back-and-forth voice conversations, typed input, and a dedicated app that references past chats for the first time.
Task execution is also more reliable. Siri consistently plays the requested song, podcast, or audiobook on the user's preferred app, which had been a persistent failure point in prior versions. It launches websites, opens directions, drafts emails, edits photos, and answers location-aware questions like when a specific artist is playing nearby. In the camera viewfinder, it identifies real-world objects and splits a restaurant bill.
Brainstorming works as a general-purpose chatbot. Siri will refine an idea, explain a new topic, or suggest a recipe from whatever ingredients are in the fridge, rather than punting the query to a web search as older versions routinely did. It is, functionally, the assistant Apple demoed at WWDC and then delayed.
The technical foundation is a partnership with Google. Apple used Gemini to train and refine its own Apple Foundation Models, which run on Apple silicon and on Apple's private cloud compute infrastructure rather than on Google's servers. The arrangement lets Apple keep the on-device and privacy-cloud story intact while borrowing the model capabilities it could not build in-house on time.
The awkward part is timing. Being a functional chatbot in 2026 is table stakes, not a milestone. Rivals have moved on to agents that write software, complete multi-step tasks, use the computer, and generate media end-to-end. Anthropic released Claude Opus 5 recently, and the broader industry is measuring assistants on autonomous work, not on whether they can find a receipt.
That gap shows up in how the launch has been received. Sarah Perez at TechCrunch, who has covered Apple since August 2011, called the new Siri exactly what users always hoped it would be — and also noted it feels like Apple fixed a long-standing bug rather than delivering something new. The marvel, she wrote, is that after years of failing, it works.
There is a business question underneath the technical one. Apple's CEO Tim Cook has floated the idea of an iCloud+ paywall for heavier Siri AI usage, which AI Chat Daily covered earlier this month, suggesting Apple sees on-device AI as a subscription lever rather than a free upgrade. Pricing the Gemini-trained backend against Google's, OpenAI's, and Anthropic's consumer offerings will get complicated fast if compute costs scale with usage.
For the general public, though, none of that matters until September, when iOS 27 ships officially and hundreds of millions of iPhones get an assistant that actually does what the icon has implied since 2011. That distribution is Apple's real weapon. A merely competent assistant on every iPhone reaches a user base that no standalone AI app has come close to touching.
Apple's Siri story now splits into two arcs. The first is the catch-up arc, which iOS 27 closes. The second is the agent arc, where competitors are already selling autonomous workflows and Apple has not shown its hand. If the Gemini-trained Foundation Models are the platform for that next step, the partnership with Google becomes structural rather than transitional, and Apple's AI independence claim gets thinner. If Apple can pivot the same infrastructure into on-device agents that act across apps, the delayed Siri launch will read in hindsight as the boring but necessary first floor of a much taller building.
Working on something we should cover, or seeing a story we missed? Send leads, documents, or feedback to hello@aichatdaily.com. For sensitive tips, see our secure tips page for Signal and PGP options.
Spotted an error? Email hello@aichatdaily.com with the URL and the issue, or read our full corrections policy.
