Apple Watch Series 12 and Ultra 4 add always-listening Audio Intelligence in a hardware-isolated Secure Exclave, drawing consent criticism from the EFF

Announced September 9, 2026, Apple Watch Series 12 ($399) and Ultra 4 ($499) add four opt-in Audio Intelligence features — Live Rewind, Siri Recap, Sound Recognition, and Music Recognition (Shazam) — that process microphone audio in a hardware-isolated Secure…

At its September 9, 2026 fall event — its first under new CEO John Ternus, which also brought the $1,999 foldable iPhone Duo (7.6-inch inner display, pre-orders October 16, availability October 23), iPhone 18 Pro from $1,199, AirPods 5 from $129, and a redesigned Health app with Apple Intelligence Insights, Readiness, and Health Age features — Apple announced Apple Watch Series 12 ($399) and Ultra 4 ($499) with four opt-in Audio Intelligence features: Sound Recognition, Music Recognition (Shazam), Siri Recap, and Live Rewind. Live Rewind transcribes the previous 15 seconds of audio when the user deliberately double-presses the Digital Crown each time, sounding an audible chime and visual cue for bystanders; Siri Recap generates titled conversation summaries and key points from ambient listening, viewable in the Siri app. Per a privacy document Apple published alongside the event, microphone audio is processed in the Secure Exclave of the S11 chip, never saved as a file, and inaccessible to watchOS, apps, the user, or Apple, with untransferred audio automatically deleted. Audio is routed to a paired iPhone, where on-device speech recognition produces a distilled transcript and foundation models in Private Cloud Compute generate summaries; watch-to-iPhone transfers are encrypted between Secure Exclaves, transcripts sync end-to-end encrypted when a device passcode and iCloud two-factor authentication are enabled, and speakers are not identified. The features also alert deaf or hard-of-hearing users to sounds such as sirens, alarms, doorbells, and crying babies. TechCrunch argued the features normalize always-listening technology and raised consent and evidentiary questions about text-only transcripts in court, noting competitors including Friend, Amazon's Bee, and Plaud in AI transcription. The Register reported EFF criticism that 11 US states require all-party consent for recording, that bystanders have no practical way to opt in or decline, and urged users to consider bystanders' conversational privacy before using the features.

• Apple Watch Series 12 ($399) and Ultra 4 ($499) were announced September 9, 2026 at Apple's fall event, its first under new CEO John Ternus.
• Four opt-in Audio Intelligence features ship on the watches: Sound Recognition, Music Recognition (Shazam), Siri Recap, and Live Rewind.
• Live Rewind transcribes the previous 15 seconds of nearby audio when the user deliberately double-presses the Digital Crown, requiring activation each time; an audible chime and visual cue alert bystanders.
• Siri Recap generates titled AI summaries and key points from daily conversations without retaining raw audio or attributing speakers, viewable in the Siri app.
• Microphone audio is processed in the Secure Exclave of the S11 chip, held in isolated memory, never saved as a file, and inaccessible to watchOS, apps, the user, or Apple; untransferred audio is automatically deleted.
• Audio is routed to a paired iPhone, where on-device speech recognition produces a distilled transcript, and foundation models in Private Cloud Compute generate the summaries.
• Watch-to-iPhone transfers are encrypted between Secure Exclaves; transcripts sync end-to-end encrypted when a device passcode and iCloud two-factor authentication are enabled.
• Apple states speakers are not identified in transcripts.