Apple Watch Series 12 and Ultra 4 Add AI Audio Features With On-Device Privacy Protections

Apple announced in September 2026 that its new Apple Watch Series 12 and Apple Watch Ultra 4 will include four opt-in “audio intelligence” tools that continuously use the devices’ microphones to power features such as sound recognition, music identification, conversation recaps, and a 15-second audio rewind function.

The features — Sound Recognition, a Shazam integration, Siri Recap, and Live Rewind — are all opt-in and built around Apple’s existing privacy infrastructure. Apple stated in a report shared with WIRED that the tools “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.”

Central to the privacy architecture is a new hardware component in the watches’ S11 chips called the Secure Exclave, a memory-protected, isolated buffer where audio data is held and processed separately from the rest of the operating system. Sound Recognition, for example, alerts users to environmental sounds — doorbells, sirens, alarms, or a baby crying — without sending any data off the device.

Features that do require cloud processing, such as Siri Recap, follow a multi-step pipeline. Audio is encrypted on the watch, transferred via secure Bluetooth to the Secure Exclave on a paired iPhone, transcribed locally, stripped of nonessential words, screened by a safety model to remove potentially harmful terms, and only then sent — as a distilled text summary, not raw audio — to Apple’s Private Cloud Compute infrastructure. Raw audio is deleted from both devices immediately after each processing step.

The Shazam feature similarly generates a song “signature” stored in the Secure Exclave rather than an audio file. That signature is sent to Shazam’s servers only if the user actively requests identification, and is deleted immediately regardless of whether the user acts.

Apple’s rollout of these features reflects the broader expansion of AI-powered audio tools into everyday consumer devices, raising questions about what persistent microphone access means in practice — even when technical safeguards are in place.

Source: WIRED