watchOS 27.3 beta brings always-on Audio Intelligence to Siri on Apple Watch

Siri gets ears on the wrist

Apple has begun rolling out a watchOS 27.3 developer beta that gives Siri two new Audio Intelligence capabilities on the Apple Watch, according to The Verge. The features arrived with the beta on November 3rd and turn the watch's microphone into the input for an always-on AI layer: one feature summarises entire conversations after the fact, while the other replays the last few seconds on demand.

The first, Siri Recap, keeps an ear on nearby conversation and, once a talk wraps up, has Apple Intelligence automatically generate a title and a set of key points inside the new Siri app. Apple's pitch, as reported by The Verge, is that users can stay present and engaged in the moment and consult the recap afterwards to jog their memory.

The second, Live Rewind, targets the moments when something was simply missed. A double-press of the Digital Crown pulls up a text transcript of the previous 15 seconds of audio detected by the watch's microphone, letting a user re-read a name, number, or phrase that went by too quickly. Users can also ask Siri questions about that text or save the snippet to the Siri app to revisit later.

A redesigned Health app rides along

The same release cycle brings a redesigned Health app into testing via the iOS 27.3 developer beta. According to The Verge, the new version adds an Insights tab that surfaces summaries of heart, sleep, fitness, and other health data throughout the day, a For You section with personalised recommendations, and educational content such as articles and videos.

Apple also announced expanded language support for the redesign, covering global English, Simplified and Traditional Chinese, French, Japanese, Korean, Norwegian, Portuguese, Spanish, Turkish, and Vietnamese.

Privacy claims underpin an always-listening device

A wrist-worn microphone that listens continuously is an obvious privacy flashpoint, and Apple has published an Audio Intelligence privacy overview to accompany the feature. As described by The Verge, Apple claims that raw audio is processed within dedicated hardware, is never saved as a file, and is inaccessible to the operating system, to apps, and to Apple itself.

Saved content is treated differently from the transient audio. Apple says that if a user has iCloud with two-factor authentication and a device passcode enabled, any Siri Recap summary or Live Rewind snippet they choose to keep syncs across their devices in the Siri app with end-to-end encryption.

Why it matters

This is one of the clearest signals yet of AI assistance shifting from on-demand cloud queries to ambient, on-device capability woven into everyday hardware. Siri Recap and Live Rewind effectively turn the Apple Watch into a memory aid, a use case with genuine utility for recalling names, instructions, or details from conversations, but also one that normalises a microphone that is always listening.

Apple's answer is architectural: processing confined to dedicated silicon, no audio files, and end-to-end encryption for anything the user opts to save. Whether those assurances satisfy regulators and users will shape how quickly always-on listening spreads to other devices and competitors. For now, the features live in a developer beta, so broader availability, final behaviour, and any changes before public release remain to be seen.