Apple shipped its rebuilt Siri to iPhone, Mac and Vision Pro owners on Oct. 26, and the release notes led with the detail that matters least to reviewers and most to the company’s growth: five more languages.
The update, iOS 27.2 on phones, macOS 27.2 on Macs and visionOS 27.2 on the headset, is the first public build in which the assistant keeps track of what was said earlier in a conversation and draws on a wider set of personal data to answer. Messages gains suggestions built on the same system. Apple did not attach a headline feature to the release, and did not put an executive in front of a camera to explain it.
That restraint is the story. Siri’s reconstruction was announced at the developer conference in June 2024 with a demonstration of an assistant that could read what was on screen and act inside apps. The company then missed its own dates, confirmed a delay in March 2025, and spent the following year absorbing criticism that it had sold a capability it could not ship. A $95 million settlement resolved claims that Siri had captured private conversations without being asked to.
What arrives now is the part that does not photograph well: language coverage. Apple Intelligence debuted in American English in October 2024. Localized English followed in December, then a batch of French, German, Italian, Portuguese, Spanish, Japanese, Korean and Chinese releases in the spring of 2025. Each wave has been a set of models, voices and evaluation suites rather than a translation of an interface.
That is why adding five languages is expensive. Apple runs a foundation model of roughly three billion parameters on the device, sized to fit in the memory and thermal budget of a phone, with heavier requests routed to servers the company calls Private Cloud Compute. A model tuned on English corpora does not become a good Japanese assistant by adding a dictionary. It needs speech recognition that handles local accents, a voice, and tests for the things people actually ask in that language, which differ by market.
The commercial logic is straightforward. Siri is the interface Apple wants between its customers and third-party apps, the piece that turns a phone into something you talk to instead of tap. That proposition only works in the languages people speak. Apple’s handset share is concentrated in English-speaking markets and Japan, and every market outside that base is one where an assistant that only understands English is a feature the buyer ignores.
It is also a market where Apple is behind. Google’s Gemini has been multilingual from the start and runs on Android, which gives it volume and language data. Amazon’s Alexa is in the home in a dozen languages. OpenAI’s voice mode made conversational latency a consumer expectation, and Samsung ships its own assistant in markets where Apple’s is thin. Siri’s reputation for misunderstanding has outlasted three rebuilds.
The regulatory map complicates the rollout. Apple held back Apple Intelligence in the European Union for months, citing the Digital Markets Act’s interoperability requirements, before shipping features there in 2025. China requires a locally approved model partner before foreign assistants can run, which is a separate approval track with its own timetable. A language list is therefore also a list of jurisdictions where the company has cleared its paperwork.
Developers watch the language list for a different reason. The rebuilt Siri acts through app intents, the system Apple built for apps to expose their functions to the assistant. An intent that is tested in English and untested in Korean produces an assistant that fails quietly in Seoul, and failures in an assistant are attributed to the assistant rather than to the app.
Apple’s own framing has shifted toward reliability over novelty. Executives have described the work as making Siri useful for the tasks people repeat, rather than demonstrating capabilities in a keynote. The features in 27.2 follow that: context from earlier in a conversation, more personal data, suggestions in Messages. None of it is a demo. All of it requires the model to be right often enough that people stop checking.
There is a version of this release that matters more than five languages. If the assistant now holds a thread across turns and reaches into messages and calendars, Apple has shipped the piece of the 2024 demonstration that users could feel, and the language expansion is the scale-up. If it has shipped a better Siri that still loses the thread on the third question, the language list is decoration.
The early signal will come from markets Apple does not discuss much. Voice assistants are used differently where typing is awkward, where data plans are metered, and where the dominant apps are not American. Those are the markets where a competent multilingual Siri changes a purchase decision, and where the company’s growth has been hardest to find.
Apple’s next earnings call will be about services revenue and handset demand rather than about Siri. Language support will not appear in a slide. It will appear, if it appears at all, in the retention numbers of the next two quarters, in markets where the phone’s main selling point has been the thing it could not do in the local language.
For now the update is live in five more languages, on three platforms, with no announcement video and no executive interview. For a company that once built a campaign around the assistant’s personality, shipping the fix quietly is its own admission of how far the promise had drifted from the product.
