Apple Watch's Live Rewind has to listen continuously to work

For a watch to tell you what someone said 15 seconds ago, it has to have been listening 15 seconds ago.

How the feature works

Live Rewind arrives with Apple Watch Series 12 and Apple Watch Ultra 4. A double-press of the Digital Crown returns a text transcript of the last 15 seconds of nearby speech. Because it runs after the words were spoken, the watch has to be continuously processing the sound around it rather than starting on command.

Apple’s answer is a set of protections. The Audio Intelligence features do not generate audio recordings. The feature activates only on a deliberate double-press. When it does, the watch sounds an audible chime from its speaker, and that chime plays even when the watch is silenced or headphones are connected, which is what prevents covert use.

Reactions split immediately, between people calling the feature antisocial and people pointing out that it is a real accessibility gain for deaf and autistic users.

What actually shifts

Both reactions can be correct at once, and that is the more useful way to read it. The privacy engineering here is substantive, and most always-listening consumer products ship with nothing comparable.

What the launch changes is the baseline. A wrist device that processes ambient speech by default is now an ordinary consumer product, and the next company to ship one is under no obligation to bring the on-device protections or the audible chime with it. A protection that cannot be verified from outside the device is a commitment rather than a guarantee, which is why the argument is about the category rather than about this particular watch.