You could be discussing something personal with an Apple Watch wearer without realising artificial intelligence is processing your words.
Apple's Live Rewind and Siri Recap features entered developer beta testing on 3 November 2026, allowing compatible watches to retrieve recent speech and generate conversation summaries.
The tools are optional, and Apple says they do not create or store audio recordings. Yet they raise a different question: will everyone involved know when their words are being processed?
Introduced through the watchOS 27.3 developer beta, the features work on Apple Watch Series 12 and Ultra 4 when paired with a compatible iPhone.
How Apple Watch Remembers Conversations
Live Rewind displays a text snippet of the previous 15 seconds of speech when the wearer double-presses the Digital Crown. It could help someone recall a name, direction or instruction.
According to Apple's support documentation, the snippet disappears approximately 30 seconds after the display dims unless saved or sent to Siri. Live Rewind does not continue transcribing in the background.
Siri Recap works differently. Once enabled, it can run at scheduled times or locations and produce brief summaries rather than complete transcripts. Users can also activate it manually. Unsaved summaries can be set to disappear within seven days.
Apple says Siri Recap prioritises conversations involving the wearer, although it may also summarise a lecturer speaking in a quiet room.
What If Other People Haven't Agreed?
During a workplace meeting or personal discussion, an Apple Watch wearer could later review an AI-generated summary. Other participants might not know their speech was processed.
Apple provides different notifications for the two features. Live Rewind plays an audible chime, even in silent mode, accompanied by an animation and microphone indicator.
For Siri Recap, Apple says the watch notifies its wearer when scheduled summarisation starts or stops. Its published guidance does not describe an equivalent audible alert to everyone nearby whenever the feature operates.
That distinction raises questions about awareness and consent. However, it does not establish that Apple or its users have violated privacy law. Legal requirements depend on the jurisdiction and circumstances.
Apple Says Audio Isn't Stored
Apple says audio enters the S11 chip's Secure Exclave, a hardware-isolated environment, and is deleted after processing. The company says operating systems, applications and Apple itself cannot access the raw audio.
For Siri Recap, encrypted audio travels to the paired iPhone for transcription and condensation inside another protected environment. Condensed text then reaches Apple's Private Cloud Compute system to generate summaries. Apple says the cloud system does not retain the information.
Saved snippets and summaries can sync through iCloud with end-to-end encryption when the required account and device protections are enabled.
Neither feature attributes speech to individual speakers, although mentioned names may appear. Apple also acknowledges that automated filters can miss sensitive information and AI-generated summaries may contain errors.
Privacy Protections Leave a Consent Question
The features require an iPhone 16 or later, Apple Intelligence and Siri AI. Initially available in English, they are unavailable to users under 18 and are not initially offered in the EU or China.
Apple advises users to consider others during private discussions. Its safeguards address how information is handled, but cannot guarantee that everyone nearby knows their words may become part of an AI-generated summary.
