Apple defends new Apple Watch's always-listening features

Apple published a technical privacy paper explaining how the always-on microphone features in the new Apple Watch Series 12 and Watch Ultra 4 process what they hear. The pitch, laid out in a document Apple shared with Wired, is that the audio never leaves an isolated corner of the S11 chip.

The suite, called Audio Intelligence, groups four opt-in tools: Sound Recognition, automatic Music Recognition with Shazam, Live Rewind, and Siri Recap. Live Rewind holds a rolling 15-second buffer of ambient audio and, when a wearer double-presses the Digital Crown, transcribes those seconds as text. Siri Recap listens ambiently and produces "a high-level summary" of conversations rather than a verbatim transcript, and Apple says it is designed to exclude financial information and government-issued identifiers.

The mechanical claim is that microphone audio flows into a buffer inside a Secure Exclave on the S11 chip, is continuously overwritten, and never becomes accessible elsewhere. "These features do not create or store audio recordings, and raw audio used for processing is completely inaccessible to the operating systems, apps, the user, or Apple," Apple said in the paper.

When Live Rewind is triggered, the Watch plays a chime even on silent and shows a microphone indicator with a full-screen animation on the display. Audio Intelligence does not attribute speech to individual speakers, though a Siri Recap summary may contain names if they are spoken aloud.

Users can leave Audio Intelligence always on, set a schedule, or turn it off. Two of the researchers we track shared the Wired write-up the day it ran, one small sign of how quickly the "always-listening wearable" framing is spreading as Meta pushes its own AI glasses.