Apple’s new Audio Intelligence turns the Watch into a listening device, privacy is part of the pitch

A hot potato: One of the main selling points of Apple’s recently unveiled Watch Series 12 and Ultra 4 is a new Audio Intelligence feature that records and analyzes ambient speech, music, and other sounds. Seemingly anticipating the concerns that come with turning a smartwatch into a listening device, the company also outlined numerous privacy protections, but it remains to be seen how people will use Siri’s new capabilities.

Pre-orders are now open for the Apple Watch Series 12 (starting at $399) and the Apple Watch Ultra 4 ($799), with availability starting on September 18. The new wearables introduce improved health-tracking features and can use Siri’s upcoming AI enhancements to analyze and transcribe what they hear.

Apple’s new Audio Intelligence includes four functions for the Watch Series 12 and Ultra 4, two of which deal directly with the user’s conversations.

When the wearer double-presses the Digital Crown, Live Rewind displays a text snippet of what was said over the past 15 seconds, which users can save to the Siri app. Meanwhile, Siri Recap listens to conversations ambiently throughout the day to generate a summary later. Both features are set to enter beta in late 2026.

The other two features seem relatively benign. Sound Recognition alerts hard-of-hearing wearers to important noises such as sirens, alarms, doorbells, and crying children. Additionally, Music Recognition automatically uses Shazam to display a song’s title and artist in the Smart Stack, without requiring users to tap the widget.

In a support article and a privacy document, Apple stresses that it processes audio in a secure part of the watch and iPhone that’s inaccessible to the user and the company, so users can only retrieve text snippets or summaries, never the underlying audio. Live Rewind constantly overwrites old audio as new data streams in, and Siri Recap deletes audio after transcription.

Furthermore, Siri Recap condenses and encrypts transcripts to retain only a conversation’s core meaning before sending them to Apple’s Private Cloud Compute servers, where they remain inaccessible to the company, to create a summary.

Siri Recap automatically deletes summaries after seven days unless users save them. It also attempts to omit financial information, harmful content, government-issued IDs, and certain other identifying data.

Even if audio data and transcripts remain secure, predicting how people will use the new listening features will be difficult. Furthermore, deleting master audio data makes it harder to verify an AI summary’s accuracy.

The Siri AI beta will roll out in English when watchOS 27 launches on September 14, with support for French, Japanese, Korean, Portuguese, and Spanish coming in October.

Compared with prior models, the Apple Watch Series 12 and Ultra 4 will also monitor heart rate and heart rate variability more accurately, and the Ultra 4 features improved battery life.