Product1 publisher2 min readPublished
Apple's 15-second Watch buffer moves the consent problem onto the person standing nearby
Apple's new Watches transcribe a rolling 15 seconds of nearby speech and summarise the day, with an audible chime as the notice to everyone else in the room. Teams copying the pattern inherit two-party consent law.
The Product Desk · Product desk
What happened
- Apple announced this month that its new Watches can continuously listen to every spoken word they detect and create summaries of whatever is going on around the wearer.
- Apple says one feature, Live Rewind, uses a rolling buffer that transcribes only the last 15 seconds of speech.
- A second feature, Siri Recap, gives only high-level summaries of the day's captured audio, and Apple says neither feature identifies individual speakers.
- An audible chime warns people nearby while the watch is transcribing, which Apple offers as the guard against surreptitious eavesdropping.
Compiled by The Product DeskSomething wrong?How this is made
Why it matters
- constraint Notice that arrives at the same moment as capture leaves no interval in which a person nearby can decline, so a feature built this way holds consent from the wearer alone.
- exposure The default ships from the vendor, but the eavesdropping risk attaches to whoever wore the device into a closed room in a state that requires both parties to agree.
- cost Gilliard's reading puts the cost on the person who bought nothing: the burden moves from company to customer, then to the stranger in earshot who has to notice a light or a beep.
- precedent If Apple's install base normalises always-on listening, every smaller team shipping passive audio gets to point at it in its own design review.
The rollout problem sits on one wrist in a closed room. By the time the chime tells the person across the table that transcription is happening, the buffer has already been holding their last sentence [3][5]. Apple's technical page fixes the retention window and says neither feature identifies individual speakers [4][16].
Chris Gilliard, an independent privacy researcher and the author of a forthcoming book called Luxury Surveillance, said the design does not describe consent at all [8]. "This really isn't how consent works. All of these are after the fact and require some action from the people who are already being recorded / listened to, rather than them being able to say 'no' before a recording begins," he said [10].
Teams building passive audio pitch the indicator as the consent step. In the design Apple described, notice and capture happen at the same time, so the earliest move available to a nonuser is an objection to a transcript that already exists [15]. The Verge described the arrangement as a form of surveillance in which people are made responsible for spotting subtle lights and beeps to know they are being recorded [12].
The Verge reported that experts warned of the legal risks of using the features, particularly in nonpublic spaces and in states whose eavesdropping laws require both parties to consent [2]. The story does not name those states. For a team putting ambient transcription into a sales org or a clinic, the room the device walks into decides the exposure, and it is the employee who walks in.
At scale, a per-device indicator stops being readable. Meta plans to offer 100 versions of its glasses by the end of the year [7], and one of the pairs it showed at Connect 2026 has no camera at all, alongside a slate of camera-equipped models [6]. Notice by per-device indicator asks a bystander to tell those variants apart on sight, including the ones whose capability is not visible [18]. Meta was widely mocked and sued this summer after an aggressive ad campaign for the glasses, according to The Verge [13].
Which of two moments comes first decides the exposure. One is the moment capture begins; the other is the earliest moment a person who did not buy the device can stop it. When the second falls after the first, the feature holds consent from one party, and the wearer is the one who explains that to a colleague, a patient, or a plaintiff's lawyer. The companies shipping these devices publish best-practices guides with suggestions for proper device etiquette [14].
What to watch
- Whether a state attorney general or a private plaintiff tests Live Rewind or Siri Recap under a two-party eavesdropping statute.
- Whether Apple gives administrators a control to disable ambient transcription on managed Watches, or leaves it to the wearer's settings screen.
- Whether Meta's 100 planned glasses variants converge on one recognisable indicator or each model signals capture differently.