Apple details how new ambient listening features keep audio private
Apple has released a document explaining how its new Audio Intelligence features process sound without ever storing or accessing users' raw audio. The company says the data is handled in an isolated hardware compartment inaccessible to the operating system, apps, or Apple itself.
At its recent event unveiling the iPhone Duo, Apple introduced several new Siri Audio Intelligence features — Siri Recap, Live Rewind, Sound Recognition, and Music Recognition. Alongside the announcement, Apple published a document describing how it intends to balance AI-driven "ambient listening" with user privacy.
According to the document, raw audio captured by these features is processed within dedicated hardware, is never saved as a file, and cannot be accessed by the operating system, apps, or Apple.
The role of the Secure Exclave
Apple explains that Audio Intelligence depends on a component called the Secure Exclave — a hardware-isolated compartment built into the S11 chip that powers the new Apple Watch Series 12 and Apple Watch Ultra 4. This compartment processes sensor data separately from the rest of the system, and none of that processed data can be accessed by watchOS, apps, the user, or Apple.
Audio picked up by the microphone enters the Secure Exclave, where it is initially analyzed to detect speech, sounds, or music, without being transcribed or stored as raw audio. Apple describes this as a continuously overwritten stream that exists only within protected hardware and never results in an actual audio recording.
User control and encryption
Apple stresses that users decide whether and when Audio Intelligence features are active. Live Rewind, for instance, must be manually triggered each time by double-tapping the Digital Crown. When data needs to move from the Watch to an iPhone, it is transferred with encryption through both devices' Secure Exclaves.
Users also choose which text generated by Siri Recap and Live Rewind to keep. If a device has a passcode set and iCloud two-factor authentication enabled, text produced by Audio Intelligence syncs with end-to-end encryption.
