PlatformsThe briefing
Apple says Siri AI began rolling out on September 14 across its OS 27 releases: English only, opt-in, absent from EU iPhone, iPad and Watch, unavailable in mainland China, and capped daily on server-backed features.
On September 14, 2026, Apple released a rebuilt Siri, which it calls Siri AI. It arrives with iOS 27, iPadOS 27, macOS 27, watchOS 27 and visionOS 27. Apple says Siri AI can use your personal context, such as your messages, emails and photos, understand what is on your screen and take actions inside apps. Apple labels it a beta.
Not everyone gets it. Apple says it starts in English, with French, Japanese, Korean, Portuguese and Spanish coming the next month. It is not available at launch on iPhone, iPad or Apple Watch in the EU, and it is not available in China while Apple works through regulatory requirements. It needs a supported device, which starts at iPhone 15 Pro, Macs with an M1 chip and Apple Watch Series 9.
Apple describes how it works. A model runs on your device, and harder requests go to Apple servers called Private Cloud Compute. Apple says the new models were built with Google and its Gemini models. Apple says that when Private Cloud Compute handles a request, your personal data is not stored and is not accessible to Apple or anyone else. That is Apple's own description. The announcement does not include an independent audit.
There are limits and a future fee. Apple says features that use server-side models have daily usage limits that vary by feature, request and demand. It says wider access will be available for a fee in the future. It gives no price and no date.
If you build an iPhone or Mac app, consider exposing your app's actions to Siri now. It is cheap and easy to undo. Do not make Siri the only way to use a feature. Plan for the people who will not have it: EU users on iPhone, iPad and Watch, users in China, users who speak another language and anyone on older devices. Treat a refusal or a fallback as a normal state your app handles. Apple does not publish what the daily limits are.
Check who is eligible before you plan around it. A company policy that assumes every employee gets the same assistant is wrong on day one, because of region, language and device. If you work in a regulated field, the data claim is Apple's statement about its own system. It is not a certificate you can hand to an auditor. Ask for it in writing and check what your device management tools can turn off. Apple says a paid tier is coming and has published no price, so any budget that assumes one is guessing.
Pick one task your staff might try by voice, such as finding a document or drafting a reply. On a test device in each region where you work, check whether Siri AI is offered at all. Then run the task ten times and write down the OS version, region, language, whether it needed the network and what happened when it declined. Your own log is the only way to learn what the daily limit allows.
Source material behind this briefing. Our practical implications are Ashlr’s interpretation.
This briefing was prepared with AI assistance from the linked sources. The analysis may include interpretation; consult the original sources for complete context. Cover artwork is illustrative.
Andy Jassy told analysts that frontier model competition will produce several comparably capable systems. Bedrock already routes to more than 100 models through one API, and most traffic now runs on Trainium.
PlatformsThe briefing
AI-generated illustration
Mistral AI has made the weights of its 1-trillion-parameter multimodal model available under open release, enabling sovereign control and local deployment options.
PlatformsThe briefing
AI-generated illustration
AWS released general availability of training sessions in SageMaker AI serverless model customization. Builders can now write their own post-training loops, including custom loss functions and multi-turn RL, while SageMaker handles GPU allocation, distributed execution, and fault recovery.
