Apple explains how new Audio Intelligence features process sound without storing raw audio
Apple published a technical document on September 9th detailing how its new Siri Audio Intelligence features - including Siri Recap, Live Rewind, Sound Recognition, and Music Recognition - process microphone audio inside dedicated hardware that neither the operating system nor Apple itself can access.
👁 115 reads
Apple published a technical privacy document on September 9th explaining how the new Siri Audio Intelligence features announced at its iPhone Duo launch event handle microphone data - and asserting that raw audio from those features is never accessible to the operating system, third-party apps, or Apple itself.
- Raw audio from Audio Intelligence features is processed inside the S11 chip's Secure Exclave and is never saved as a file or accessible to watchOS, apps, or Apple
- Live Rewind requires a manual double-tap of the Digital Crown each time a user wants to activate it
- Audio Intelligence data transferred to iPhone travels with encryption through both devices' Secure Exclaves, and text syncs with end-to-end encryption if a device passcode and iCloud two-factor authentication are enabled
The four features - Siri Recap, Live Rewind, Sound Recognition, and Music Recognition - rely on a hardware component called the Secure Exclave, built into the S11 chip that powers the Apple Watch Series 12 and Apple Watch Ultra 4. According to Apple’s document, audio from the microphone enters the Secure Exclave where it is processed for speech, sounds, or music without transcribing or storing the raw audio. The buffer, Apple says, is a continuously overwritten stream that exists only within the protected hardware and never creates an audio recording.
Apple describes the Secure Exclave as “a hardware-isolated compartment built into the silicon to process sensor data separately from the rest of the system,” adding that data analyzed there “cannot be accessed by watchOS, apps, the user, or Apple.”
Google · Your sourcesQuantslant higher in your Google - one clickAdd to preferred sources →User controls and data transfer
Apple frames the new features as opt-in by design. Live Rewind, for example, requires a user to double-tap the Digital Crown every time they want to activate it - it does not run continuously in the background without a deliberate action. Apple says users control “whether and when” Audio Intelligence features are active.
When Audio Intelligence output needs to move from an Apple Watch to a paired iPhone, the transfer is encrypted through the Secure Exclaves of both devices. Users also choose which text to keep from Siri Recap and Live Rewind transcripts. If a user has a device passcode set and uses iCloud with two-factor authentication enabled, any text that syncs across devices does so with end-to-end encryption.
The timing of the document - released alongside the product announcement rather than after public scrutiny - reflects a pattern Apple has followed with other sensitive AI features, where technical privacy details are published proactively to get ahead of questions about ambient listening. Whether the architecture delivers on those assurances in practice is something security researchers are likely to examine once the Apple Watch Series 12 and Ultra 4 ship to users.
More on this from QUANTSLANT
magazine·OpenAI Claims Navier-Stokes Breakthrough, but Accusations of Academic Scooping Follow