Siri Recap leaves bystander consent entirely to whoever is wearing the watch

The wearer makes the choice once, in a settings screen, and then walks into a one-on-one about somebody's pay. PCMag reports the choice is between on-demand and always-on, and that it is optional only for the person wearing the watch ^[6]. Neither PCMag nor Mashable describes an indicator on the watch showing the other party which mode is running ^[22]. The evidence does not show always-on shipping as the default, only a setting one person picks and nobody else sees ^[6].

Apple's privacy claim is narrow, and the narrowness is what matters to anyone drafting a rule. Apple says Siri Recap will not upload audio to the cloud or store it on the device, and will not identify speakers by name in the summary ^[8]. PCMag describes the path: the watch sends captured audio to the iPhone, the phone transcribes it, and the phone then sends a condensed version of the text to Apple's Private Cloud Compute servers, where the summary is generated ^[7]. The third of those stages runs off the wearer's hardware ^[19].

The vocabulary differs between the two accounts. Mashable reports Apple describes the raw audio data as notes and not exact transcripts, and says it is not accessible by app developers, users, or Apple itself ^[10]. The same story reports Apple calling recorded conversations end-to-end encrypted while also saying devices do not collect or store actual audio recordings ^[9]. PCMag calls the thing that leaves the phone a text transcript ^[7].

That distinction lands squarely on statutes written about recording. PCMag notes that many US states require two-party consent, and that Siri Recap might not meet the legal definition because it converts audio to transcripts instead of permanently storing recordings; the publication calls that a loophole running counter to the spirit of those laws ^[13][14]. Its writer says plainly that he is not a lawyer ^[13].

Apple's instruction to the wearer is one sentence. "While audio is not recorded and stored, and no speakers are attributed, consider those around you where conversations might be private or sensitive," Apple wrote ^[12]. Apple did not elaborate on the guardrails behind it during Wednesday's keynote ^[18], though its blog post says recaps omit self-harm and hate speech and filter out financial information and authentication data ^[11].

Teams tell themselves the existing recording policy covers this, because it says no recording without consent. The wearer never files Recap under recording, because the vendor does not describe it that way ^[10]. The summary shows up without anyone else in the conversation having agreed to anything ^[20]. So the rule has to name capture and summarisation of speech, not recordings, to reach this feature.

The questions that decide any given room are whether the other party can tell capture is running and whether any product of that capture leaves the wearer's device. With Siri Recap, the other party cannot tell, and a product of that capture does leave ^[22][7]. Rooms where you already ask a visitor to put a phone face down need the same sentence written for wrist wearables. PCMag notes many people wear the Apple Watch almost 24/7, including people handling sensitive information in education, finance and healthcare settings ^[15]. Live Rewind with Siri Recap goes into beta this fall, starting in English ^[5].