Apple details how Apple Watch’s new Audio Intelligence features work in privacy paper

Apple has unveiled a suite of new features known as Audio Intelligence, set to debut on the upcoming Apple Watch Series 12 and Apple Watch Ultra 4. The company’s white paper outlines the privacy measures in place for these innovative features, which include Sound Recognition, automatic Music Recognition with Shazam, Live Rewind, and Siri Recap.

Audio Intelligence utilizes Apple's microphones, on-device and cloud-based AI models, operating with stringent privacy safeguards. Live Rewind and Siri Recap will initially launch as beta features in English later this year, with more information forthcoming upon official release. Users have the option to individually enable or disable these features as needed.

Live Rewind only activates upon double-clicking the Digital Crown and displays a 15-second text summary of the last spoken words, without continuous transcription or background saving. Siri Recap runs at the user’s discretion, with customizable schedules based on time, location, or manual activation. The privacy paper explains that Live Rewind's audio data flows into a secure buffer within the S11 chip, preventing accumulation and unauthorized access.

Siri Recap's audio detection is performed by a lightweight AI model that doesn't transcribe or record conversations, only determining if speech is occurring nearby. The model's audio buffer remains isolated and inaccessible to users, the operating system, apps, or Apple. For a comprehensive understanding of the new features and their privacy measures, the full 11-page paper is available for review.