If your company runs a mix of iPhones, iPads and Macs, Apple's 14 September release brings three separate decisions to your fleet. According to Apple's announcement, Siri AI is not initially available in the EU on iOS, iPadOS and watchOS, while it is available on Mac and Vision Pro. The practical consequence follows from that: do not build a workflow on the phone assistant, allow narrow Mac-side testing, and keep personal-context and onscreen-awareness features off until you have read through the data handling and device management settings. For a Hungarian-working team this release offers no usable assistant feature, and the next realistic checkpoint is the October language expansion.

What actually shipped on 14 September

Apple began rolling out iOS 27, iPadOS 27, macOS 27, watchOS 27, visionOS 27 and tvOS 27 that day, together with Siri AI, the next generation of Apple Intelligence. As the two newsroom pages describe it, the new assistant understands personal context across messages, emails and photos, can answer questions about onscreen content, and performs more systemwide app actions. Alongside it came new Photos editing tools (Spatial Reframing, Extend, an upgraded Clean Up), an Image Playground that now produces photorealistic imagery, Safari tab organisation and change notifications, and description-based automation in Shortcuts.

One qualifier belongs up front. Apple calls Siri AI a beta and says the features start rolling out that day. It does not follow that every supported device has it the next morning. The original texts are here: the platform update summary and the Siri AI announcement.

The EU split inside a single fleet

The decisive sentence sits in the availability section: Siri AI will not initially be available in the EU on iOS, iPadOS and watchOS, while EU users on Mac and Apple Vision Pro can access it when the device is set to a supported language. On those same three platforms the exclusion also extends to features that rely on Siri AI. Apple's wording is that it is working to find a path forward that preserves user privacy and security. It gives no date.

PlatformSiri AI in the EU
iOS (iPhone)not available initially
iPadOSnot available initially
watchOSnot available initially
macOSavailable with a supported language setting
visionOSavailable with a supported language setting

The table follows Apple's availability statement of 14 September 2026.

For a Hungarian or Central European fleet this means two capability tiers inside the same estate. Most employees work on phones, and phones get no assistant from this release. Anyone who planned for field colleagues to dictate notes by voice, pull a figure back out of an old email or trigger an app action has no EU deployment path today. Mac availability is fine for trials and falls short for fleet-level process design.

The same section states that Siri AI and the other new Apple Intelligence features will not be available in China while regulatory requirements are worked through. Jurisdiction shapes this rollout alongside hardware eligibility.

What probably still works on EU iPhones

This deserves a precise reading, because the exclusion does not cover everything. Apple describes the app-level Apple Intelligence capabilities as available on supported products set to a supported language, and states the territorial exclusion specifically for Siri AI and features that rely on it on iOS, iPadOS and watchOS. That implies Photos editing tools, Image Playground or Safari tab organisation remain available on an EU iPhone when the device language is on the supported list. Apple does not enumerate this, so in a managed fleet it has to be verified on a device before anyone promises users a feature list.

The missing Hungarian language and what follows from it

Siri AI launches as an English-language beta, with French, Japanese, Korean, Portuguese and Spanish arriving in October. Hungarian does not appear in the listed Apple Intelligence languages: English, Danish, Dutch, French, German, Italian, Norwegian, Portuguese, Spanish, Swedish, Turkish, Vietnamese, Chinese (simplified and traditional), Japanese and Korean. Hungarian-language dictation, drafting help and smart replies are therefore not part of this release. The list may expand later, which makes the October language wave the first realistic checkpoint for Hungarian use too, even though the announcement contains no Hungarian commitment.

In practice a Budapest team would need to set both the device and Siri language to English. That switches the entire system language and raises accuracy and training questions for colleagues working with Hungarian material. For mixed-language teams that already correspond in English it can work. For a team that invoices and serves customers in Hungarian it introduces a quality risk into daily work.

If your processes need throughput guarantees, work through the strategic sequence before building on any assistant. AI strategy consulting

The daily request cap you cannot plan capacity against

Both announcements carry the same sentence: Apple Intelligence features that rely on server-side models are subject to daily usage limits, including Siri AI, intelligent photo editing tools, Image Playground and cloud models in Shortcuts. Limits may vary by feature, request complexity, system demand and other factors, and expanded access will be available for a fee in the future. No number is published, and no price.

That sentence decides which bucket the technology belongs in. An illustrative example in which every number is an assumption: a forty-person team making twelve assistant requests per person per day would generate 480 server-side requests daily. Whether that sits under or over the cap cannot be determined today, because Apple publishes no quota. A process with a throughput requirement, such as clearing a daily invoice batch or preparing customer-service replies in bulk, cannot be planned under those terms. This is the personal productivity category; process automation needs a predictable quota and contractual terms.

The consequence: document processing, customer-service response drafting and compliance-sensitive summarisation stay on platforms where request limits, data processing terms and admin controls are contractually visible. We covered that choice in more detail here: Which proprietary AI model should run your enterprise system in autumn 2026?

Privacy and device management: what the announcement says and what it omits

Apple states that the new foundation models were custom-built in collaboration with Google and its Gemini models. That is relevant to supplier and sub-processor reviews, so it belongs in the vendor assessment file. The privacy assurance also comes from the vendor: when Private Cloud Compute handles requests, personal data is not stored nor made accessible to Apple or anyone else. That is a vendor statement in a press release. It is a useful input to a data protection impact assessment and does not replace one.

The conversation features that would sharply raise workplace recording questions did not ship on the day: Siri Recap and Live Rewind arrive in beta later this year. Two further constraints belong here. Apple ties these Audio Intelligence features to the S11 chip on Apple Watch Series 12 and Ultra 4, so existing Series 9 watches in a fleet will not get this set even on hardware grounds. The features also build on Siri AI, and watchOS is excluded in the EU, which implies there is nothing yet to put before employee representatives or a works council, although Apple does not state that separately. Apple's claim is that Audio Intelligence features do not create or store audio recordings, and that the raw audio used for processing is inaccessible to the operating system, apps, the user or Apple.

What neither announcement addresses: device management. There is no mention of MDM keys, admin-level disablement, data processing agreements or enterprise terms. For a managed fleet, the administrator documentation is therefore the basis for enabling personal context and onscreen awareness; the press release leaves that question open.

Where a trial makes sense today

Trial now if you run an English-working, Mac-centred team where dictation, drafting help or Safari tab organisation saves a specific colleague real time. The risk is low, the feature is actually available, and the experience will be useful later.

Hold off if the process lives on the phone, because that is exactly where the feature is missing in the EU. The same applies if your working language is Hungarian. Hold off as well if you expect throughput guarantees or calculable costs from an assistant, because both the size of the daily cap and the future pricing are unknown.

Anyone moving regulated data faces hardware eligibility as the next filter. Per Apple's list, Apple Intelligence runs on iPhone Duo, iPhone Air, iPhone 16 models or later, iPhone 15 Pro and 15 Pro Max, iPad mini (A17 Pro), M1-or-later iPads and Macs, MacBook Neo (A18 Pro), Apple Vision Pro, Apple Watch Series 9 or later and Apple Watch Ultra 2 or later, plus Apple Watch SE 3 when paired with an Apple Intelligence-enabled iPhone. Older parts of a fleet get nothing.

Two further details matter. Siri AI is not available for users under 13, which needs handling in education settings and on shared devices. And some showcased conveniences are limited to the United States from the start, such as splitting a bill with Apple Cash.

What would change this recommendation? Three things: Apple's administrator documentation publishing a figure for the daily quota, the appearance of enterprise terms and MDM keys covering Siri AI and personal context, or a date for EU availability on iOS. Until one of those happens, the sequence above holds.

Steps to take before your next fleet decision

  1. Produce a platform-by-platform count of how many devices in your fleet are eligible for Apple Intelligence at all, and put the date in the file name.
  2. Open Apple's device management documentation and write down exactly which MDM keys exist today for personal context, onscreen awareness and the Siri app. If the list is empty, that is also a documented result, with a date.
  3. Record in your vendor assessment that the models were built in collaboration with Google's Gemini models, and ask your data protection lead whether this touches the existing processing records.
  4. Run a narrow trial on Macs only, with English language settings and non-sensitive content, and measure how much time dictation and drafting help actually save.
  5. Set the October language expansion as your re-evaluation point, and until then keep document processing and customer-service replies on your contracted systems. We collected the process-level alternatives here: AI document processing in practice.