Siri on Apple Watch can remember useful parts of your day without retaining everything

Your Apple Watch now listens all day to build a daily Siri Recap from the sounds around you. The catch? It actually forgets the audio the moment it understands the context.

Audio Intelligence uses Sound Recognition to identify relevant information from the wearer's surroundings as events happen. Siri can combine those signals with other available data to summarize the day later.

Sound Recognition can identify events such as sirens, alarms, doorbells, and a crying baby, giving the system potential safety and accessibility uses. A separate Live Rewind feature lets users double-press the Digital Crown to replay the previous 15 seconds of detected audio and view a transcript.

The resulting Siri Recap doesn't include transcripts of overheard conversations, and Apple says the feature doesn't store audio. Sound processing happens continuously before the original audio is discarded.

Live Rewind temporarily exposes the audio and transcription a user requests, while Siri Recap retains only the contextual information needed for its summary. That design prevents the watch from becoming a searchable archive of everything its microphones hear.

Users can leave Audio Intelligence running throughout the day, activate it manually, or limit it to selected locations. The feature extends the AI push Apple introduced across its latest hardware.

Audio Intelligence requires Apple Watch Series 12 or Apple Watch Ultra 4, which include hardware for securely processing ambient sound. Live Rewind and Siri Recap will enter beta later this year with English support first.

Details about Siri Recaps surfaced in an earlier report, but Apple's announcement explains how Sound Recognition powers the feature. The approach brings the daily-memory concept used by dedicated AI wearables to Apple Watch without keeping a permanent audio archive.

Discarding recordings reduces the privacy risks associated with ambient listening, although users will still need to decide when continuously analyzing nearby conversations is appropriate. Location controls and manual activation give them ways to keep Audio Intelligence out of sensitive settings.