Apple Watch’s new AI features are normalizing the idea that technology is always listening

At its Surprise and Shine event, Apple unveiled several Apple Watch features that process ambient sound using on-device AI. Audio Intelligence is designed to detect environmental sounds such as alarms and sirens and notify the wearer — functionality Apple positions as helpful for people who are deaf or hard of hearing and potentially life-saving in emergencies. The company says these sound-detection features can operate even when the wearer’s iPhone is not nearby.

Other additions move beyond accessibility into broader transcription and summarization. Live Rewind lets a watch user double-press the Digital Crown to capture and transcribe up to 15 seconds of recent speech; those text snippets can be stored in a dedicated Siri app. Siri Recap takes a different approach, passively listening to create high-level notes, titles, summaries, and key points rather than verbatim transcripts, and presents those recaps in the Siri iOS app.

Apple emphasized privacy controls around the features: Siri Recap does not default to always-on, users can restrict when it runs (for example, only during work hours), and the company says raw audio is not created or stored and is inaccessible to Apple. The generated transcripts and recaps are reportedly protected with end-to-end encryption, and Apple notes visual and audible cues — an animation and chime — accompany actions like Live Rewind.

Despite those safeguards, the company’s move revives unresolved questions about consent and the social impact of subtler recording tools. Commentators cited in coverage note startups and other devices already pursuing always-available transcription and memory-assistant use cases, and legal experts say how courts treat text-only transcripts derived from ambient audio is uncertain and likely to vary by jurisdiction. Critics also point out that a quieter capture gesture on a wristwatch differs from the conspicuous act of taking a photo or video, which could change how people behave when they know conversations might be converted into text.