Is the Apple Watch listening? Since November 3, 2026, when Apple released the first developer beta of watchOS 27.3, the answer for an Apple Watch Series 12 or Apple Watch Ultra 4 with two new features switched on is yes, in a narrower sense than the question implies. Siri Recap listens for speech through the day, turns it into text and sends a condensed version to Apple's servers for a summary, while Live Rewind keeps a rolling buffer of sound so it can show the last 15 seconds as text. Apple says no recording is ever made, and its support documentation, dated the same day, states that neither feature is available to anyone under 18.

In Short

The newest Apple Watches can now listen to conversations, but only if their owner switches the features on: one shows the last 15 seconds of what was just said, and the other writes short notes about conversations during the day. Apple says the sound is turned into text and thrown away without ever being saved, and that nobody, Apple included, can hear it. The catch is for everyone else in the room, because the note-taking feature gives them no signal, and lawyers in several US states have asked whether the law still treats that as listening in.

What shipped on November 3

Apple's developer release log lists watchOS 27.3 beta, build 24S5284h, on November 3, 2026, published alongside the first betas of iOS 27.3 and iPadOS 27.3 (both build 24C5037i), macOS 27.3, tvOS 27.3, visionOS 27.3 and Xcode 27.3. A developer news post the same morning told registered developers that "The beta versions of iOS 27.3, iPadOS 27.3, macOS 27.3, tvOS 27.3, visionOS 27.3, and watchOS 27.3 are now available."

Neither entry names Siri Recap or Live Rewind. The watchOS 27.3 beta release notes deal only with StoreKit, covering one fixed bug involving subscription bundles and one known issue where some StoreKit Test calls might hang. Apple issued no press release for the beta. Confirmation that the two listening features work in the new build comes from outlets that installed it. MacRumors reported that Live Rewind and Siri Recap are available in the first beta. 9to5Mac said they can be tested by pairing the watchOS 27.3 beta with the iOS 27.3 beta, and Thurrott wrote that the first beta enables both on the two new watches.

Apple's reference page, About Audio Intelligence features on Apple Watch Series 12 and Apple Watch Ultra 4, carries a published date of November 3, 2026. The page itself is older: Apple's September 9 press releases for the two watches already linked to the same address. Its requirements section still describes Live Rewind and Siri Recap as features that "will be available in beta later this year," wording that predates the build.

Some early coverage ran ahead of the release. Engadget first published its story at 4:21 a.m. Eastern on November 3, hours before the 10:00 a.m. Pacific timestamp on Apple's release log. An update added at 9:15 a.m. Eastern said the beta was not yet available in the US and, citing TechCrunch, that Apple had announced it in India. A second update, timestamped 2:33 p.m. Eastern, said the story had been revised to note that the watchOS 27.3 beta was officially live in the US and other regions.

Apple has not said when watchOS 27.3 will reach the public. MacRumors expects it in December, alongside iOS 27.3.

Is it listening? Four features, four answers

Apple introduced its Audio Intelligence features on September 9, 2026, in the Apple Watch Series 12 and Apple Watch Ultra 4 press releases, at the same event where Apple presented the $1,999 iPhone Duo. "Since Apple Watch is always on a user's wrist, it is the perfect device to help users stay alert to important sounds and stay present and engaged in conversations throughout the day," the Series 12 release states.

There are four features, all fed by the watch microphone and all routed through the Secure Exclave on the S11 chip, which Apple describes as a hardware-isolated compartment that processes sensor data separately from the rest of the system. Each is opt-in. Apple's Audio Intelligence Privacy Overview, dated September 2026, describes the common starting point: "Audio from the microphone enters the Secure Exclave of Apple Watch, where it is initially processed for speech, sounds, or music, without transcribing or storing the raw audio. This buffer is a continuously overwritten stream that exists only within the protected hardware and never creates an audio recording."

So the watch does listen, in the sense that sound reaches a chip that analyzes it. What it listens for, and what happens next, differs sharply by feature.

Sound Recognition: listening for sounds, not words

Sound Recognition listens for sirens, alarms, doorbells and a baby crying, and sends a notification when it hears one. "Audio is buffered into small segments within the Secure Exclave and analyzed for specific sounds, then immediately deleted if nothing is detected in that buffer," the overview states. "No audio is accessible outside of the Secure Exclave on Apple Watch, and no audio is transcribed or transmitted." It is switched on from the iPhone's accessibility settings. One detail matters for anyone asking whether the watch is listening without having been told to: if Sound Recognition is already on for the iPhone, it is automatically enabled on the watch, according to the support page.

Shazam: listening for music

Music Recognition with Shazam listens for songs. When it is enabled, the Secure Exclave "is designed to detect music playing around" the wearer, according to the overview. When the Smart Stack is opened and music has been detected, the chip creates an acoustic signature, "a compact representation that cannot be reversed into the original audio," and sends that signature, not the sound, to Shazam's servers. Like Sound Recognition, it was already available on the new watches before the beta, according to MacRumors, and an existing Music Detection setting carries over without a second opt-in.

Live Rewind: a rolling 15-second window

Live Rewind shows speech that happened before the wearer asked for it. "A double press of the Digital Crown shows the previous 15 seconds of a conversation as a text snippet, so the user can catch something they may have missed or are less familiar with," according to the Series 12 release.

Showing words spoken before the press means sound has to be held before the press. The overview describes this under the heading "A continuous rolling buffer": "Audio from the microphone flows into the audio buffer inside the Secure Exclave on the S11 chip. Old audio is continuously replaced by new audio, so audio doesn't accumulate, and raw audio isn't accessible outside this protected buffer." It does not say how long the buffer is in total, or whether it runs whenever Live Rewind is enabled or only once speech is detected.

The press is what turns sound into words. "After activating, the immediately preceding 15 seconds of buffered audio is sent from the Secure Exclave of the Apple Watch to iPhone," the overview states. "If iPhone is not within wireless range of Apple Watch, the transfer will fail with an error, and the audio will be immediately discarded." On the iPhone, on-device speech recognition converts the audio to text and the raw audio is discarded. The text returns to the watch and disappears about 30 seconds after the display dims, unless the wearer saves it to the Siri app or asks Siri about it, at which point only the text goes to Private Cloud Compute, Apple's server system for Apple Intelligence. The support page acknowledges that Live Rewind "may transcribe what someone said word-for-word, but does not attribute quotes to speakers."

Live Rewind is the one feature built to tell other people it is running. "Live Rewind plays an audible chime when it is being used, even if Apple Watch is on silent, and shows a full-display animation and microphone indicator so people nearby can hear and see that the feature has been activated," the press release states. The chime plays from the watch speaker even when headphones are connected, and Macworld reported in September that Apple said it cannot be switched off. Owners can opt in during watch setup or later in Siri settings.

Siri Recap: listening through the day

Siri Recap is the feature behind the question. It does not wait for a button. "When Siri Recap is turned on, after a conversation, Apple Intelligence automatically generates a title and key points in the new Siri app, allowing users to be more present and engaged in the moment," the press release states. The output, in Apple's words, is "a brief, high-level summary, not a detailed record or transcript."

While it is on, a detector runs constantly. "The S11 chip uses a lightweight AI model to determine whether speech is occurring nearby. This model doesn't transcribe or record anything. It simply detects whether a conversation has started," according to the overview. A table in the same document lists the "Audio stream during speech detection" as processed on the "S11 Always On Processor" and accessible to "No one," with raw audio "processed on-device and discarded in real time." Once a conversation is detected, audio flows into the protected buffer and on to the iPhone.

The wearer decides when that detector runs. "They can decide exactly where and when Siri Recap takes notes, and can turn Siri Recap on or off at any time, right from Control Center," the press release states. The support page describes schedules by time, by location or both, with examples such as "only at work" or "never at night," as well as manual switching, and the watch notifies its wearer when Siri Recap starts or stops on a schedule. It pauses while the wearer is listening through headphones paired to the iPhone or watch, including during media playback, phone calls and FaceTime, and when Active Noise Cancellation is on for AirPods or Beats.

What happens to the words Siri Recap hears

The overview traces Siri Recap's path in stages.

Transfer. The watch's Secure Exclave encrypts the audio and sends it to the Secure Exclave on the paired iPhone, then deletes its own copy once the transfer completes. "If a connection cannot be re-established in time, the encryption keys expire and the encrypted audio is permanently deleted," the overview states. It does not say how long that window is.

Transcription and condensing. On the iPhone, on-device speech recognition turns the audio into text, and an on-device language model shortens it to "less than half the length of the original transcript," removing what Apple calls "content suggestive of conversation tone, non-essential language, filler words, and redundancies" while keeping the topics. "Once processed, the raw audio is immediately and permanently deleted and no longer exists on either Secure Exclave. Only the condensed text remains. Finally, a safety model screens the text to omit potentially harmful terms," according to the overview.

That sequence explains a phrase that can read as contradictory. Apple says Siri Recap does not produce a transcript, and that is true of what the wearer receives. A transcript does exist for a moment inside the iPhone's Secure Exclave, as an intermediate step, before it is condensed. The condensing works as a form of data minimisation applied before anything leaves the phone.

Cloud summarization. The condensed text is encrypted and sent to Private Cloud Compute, where Apple Foundation Models produce and format the summary. It does not travel alone. "Contextual information is also sent to improve summary quality," the overview states. Its examples are Now Playing data, so the model can tell that speech came from music or a podcast; calendar data, to generate accurate titles; "High-level location labels such as home, work, and school, along with locality information such as city, state, and country"; and point-of-interest categories such as "grocery store" or "park." "Precise location and specific points-of-interest are not included," according to the overview. The support page refers to the same step only as "some contextual information."

Return and retention. The finished summary is encrypted and sent back to the iPhone and watch, where it appears in a Recaps tab in the Siri app. There it can be viewed, edited, deleted or exported to Notes, Journal and other apps. On automatic deletion, Apple's two documents use different wording. The overview says "Siri Recaps are automatically deleted after 7 days" unless saved. The support page says they "can be set to automatically delete after a period of up to 7 days." One reads as a fixed rule; the other as a setting with a seven-day ceiling. MacRumors reported the seven-day deletion as automatic.

Some content is meant never to reach a summary. According to the support page, Siri Recap is designed to omit harmful content such as language promoting self-harm or hateful speech, along with financial data, government-assigned identifiers, authentication data and "select personal identifiers." The same page concedes that "automated filtering may not remove all sensitive information," and warns that summaries "may be incomplete or inaccurate. Important details may be omitted, misunderstood, or summarized incorrectly."

What Apple says cannot happen

Apple's case rests on one claim: listening is not recording. "There is no recording to share, forward, or produce if requested by any party, because no recording exists," the support page states. Raw audio cannot be reached by the operating system, apps, the owner or Apple, according to the same page. Saved summaries and snippets sync through iCloud with end-to-end encryption when two-factor authentication and a device passcode are in place, and "Apple does not hold the encryption keys." On the server side, data in Private Cloud Compute "is never stored or accessible by Apple or anyone else," and "independent security experts can verify this promise at any time."

Apple's leadership made the same argument in September. Chief executive John Ternus told Tom's Guide that Apple approaches such features with privacy and security in mind and treats the subject seriously. Greg Joswiak, Apple's senior vice president of worldwide marketing, told the same outlet that the absence of a recording means nobody else can obtain, hack or receive one.

The documents leave some questions open. Neither the support page nor the overview gives the length of the Live Rewind buffer or of the window in which Siri Recap audio waits to reach the iPhone. Neither says how much battery the always-on speech detection uses, or what the microphone does when no Audio Intelligence feature is enabled.

Who else is being listened to

The features have circulated online with the description that the Apple Watch listens to and analyzes all nearby conversations. Apple's documentation narrows that and also confirms part of it.

The narrowing is clear. Nothing listens until its owner opts in. Siri Recap runs only when switched on or scheduled, and stops for headphone listening and noise cancellation. It also ranks what it hears: Siri Recap "uses signals like speech duration and intelligibility to prioritize" conversations the wearer is taking part in over "conversations between people passing by," according to the support page. In noisy or group settings, Apple says, it focuses on the conversation most audible to the wearer.

The confirmation is equally clear. Once Siri Recap is on, the speech detector is listening for any speech nearby, and the speech it processes belongs to everyone in the conversation, not only the person who enabled it. Apple's own example makes the point: in an otherwise quiet lecture room, Siri Recap summarizes the lecturer. Speakers are not identified or labelled, but names spoken aloud can appear in a summary, according to the support page.

The sharpest difference between Live Rewind and Siri Recap concerns those other people. Live Rewind chimes and lights up its screen. Siri Recap does neither. "There is no audible signal because no raw audio is retained, and there is no way to reconstruct the original audio from a Siri Recap or share raw audio with anyone," the overview states. The only Siri Recap notification Apple describes goes to the wearer. Apple's guidance asks wearers to "consider those around" them "where conversations might be private or sensitive."

Whether listening without recording is still eavesdropping became a legal question within days of the announcement. Bloomberg reported on September 11 that attorneys in California, Massachusetts, Pennsylvania and Washington, states where every party to a conversation must consent to it being recorded, had questioned whether the features fit eavesdropping law, according to accounts by The Next Web on September 14 and 9to5Mac on September 15. The central argument, as 9to5Mac summarized it, is that speech processed continuously, with no conventional recording at the end, may still count as interception under state law.

The lawyers did not agree with each other. The Next Web named Joe Simons, a criminal defense lawyer in Boston; Phil Lauer, a defense attorney in Pennsylvania; Mark Blair of the Seattle firm Blair Kim Moeller; and Colin Zick, who chairs the privacy practice at Foley Hoag. Lauer suggested the technology came close to legal because no individual actually receives and uses the data, according to 9to5Mac. Blair took the opposite view, describing use without consent as a criminal violation, according to The Next Web, which added that Illinois and Florida are among about half a dozen further states with an all-party consent rule. Apple rejected the description of the features as recording conversations, The Next Web reported. iDevice, summarizing the lawyers' comments, concluded that any legal exposure would fall on the wearer rather than on Apple.

Apple's "no recording exists" line answers the evidentiary part of the question: there is nothing to hand over. Whether it settles the interception part is a different matter, and no court has ruled on it for these features.

The surrounding case law is moving. Otter.ai was sued on August 15, 2025 over claims that its Notetaker recorded meeting participants without everyone's consent. On July 30, 2026, Granola faced a seven-count complaint that includes two counts under the California Invasion of Privacy Act, which carries statutory damages of $5,000 per violation. Both cases concern services that keep recordings or transcripts and allegedly use them for model training, which Apple says Siri Recap does not do. Whether courts see that difference as decisive is untested.

Apple has been here before with a different kind of listening. At the start of 2025, Apple agreed to pay $95 million to settle a class action over Siri recordings, a case filed in August 2019 after reports that contractors reviewed audio captured when Siri was triggered unintentionally. Google followed a similar path: in January 2026, plaintiffs sought approval of a $68 million settlement over Google Assistant recordings. In both cases the complaint was about audio that existed and was heard by people. Apple's design for Siri Recap is built so that no such audio exists.

Europe and the indicator debate

Siri Recap's absence from the EU means European regulators have not yet had to apply their recent thinking on wearables to it. That thinking is specific. On October 13, 2026, the European Data Protection Board said smart glasses makers cannot rely on wearers to meet GDPR duties and pointed to LEDs visible in many lighting conditions and audio cues as possible safeguards. A month earlier, on September 10, the Hamburg data protection authority published a 53-page report finding that Ray-Ban Meta glasses expose bystanders, noting that the capture light is often too dim to notice.

Hardware makers have responded with indicators that are harder to defeat. In July 2026, Meta began disabling the camera on its second-generation AI glasses if the capture LED is destroyed. In California, Senate Bill 1130, introduced on February 17, 2026, would prohibit disabling an indicator light or other signal on a wearable device that shows it is capturing sound or video.

Live Rewind's chime, animation and microphone icon fit that model of visible and audible notice. Siri Recap, by Apple's own account, offers people nearby no signal at all, and Apple's stated reason is architectural: no audio is kept. How a European authority would weigh that argument against the EDPB's emphasis on notice at the moment of collection is an open question, and one Apple has so far avoided by keeping the features out of the EU.

Where the words are processed

Siri Recap's summaries are written by "Apple Foundation Models running on Private Cloud Compute," according to the support page. In June, when Apple unveiled Siri AI, the assistant was described as running on Apple Foundation Models built in collaboration with Google and its Gemini family, with server-side processing through Private Cloud Compute. The Audio Intelligence documents do not say whether the models that write Recaps are the same ones. That same month, Apple extended Private Cloud Compute to Google Cloud infrastructure and Nvidia GPUs, Help Net Security reported on June 10, citing Apple's security blog. Apple said the system's security model was unchanged and that it kept full control of the software wherever it runs.

Neither the support page nor the overview says whether Siri Recap requests, with their condensed conversation text and location labels, are handled in Apple's own data centers or in the Google Cloud capacity added in June.

Who can switch it on: hardware, language, region and age

The listening features require one of the two watches with the S11 chip, Apple Watch Series 12 or Apple Watch Ultra 4, which went on sale on September 18. Live Rewind and Siri Recap also need a paired iPhone from a specific list: iPhone 16, 16 Plus, 16 Pro and 16 Pro Max; iPhone 17, 17e, 17 Pro and 17 Pro Max; iPhone Air; iPhone 18 Pro and 18 Pro Max; and iPhone Duo. Both require Apple Intelligence and the Siri AI beta. Sound Recognition and Shazam detection work with a far older range, iPhone 11 or later or iPhone SE (2nd generation or later) running iOS 27.

The list includes iPhone 17e but leaves out iPhone 16e. That matches a separate line in the overview, which names "iPhone 16, iPhone 16 Pro, iPhone 17e, iPhone Air, and later models" as the iPhones with a Secure Exclave for raw audio. Apple does not state that the exclave is the reason for the exclusion.

"Siri Recap and Live Rewind will be available in English to start and will not initially be available in the EU," according to the support page's second footnote, which adds that they "are not available in China." The EU gap follows Siri AI itself, which "will not initially be available in the EU on iOS, iPadOS, and watchOS," according to the third footnote. PPC Land reported on September 14 that Siri AI shipped as a beta outside the EU and China, with Audio Intelligence for the new watches due later in the year, and that EU Macs received Siri AI while EU iPhones did not. In its November 3 developer post on Siri AI, Apple said it "is working constructively with the EU in an attempt to find a path forward."

Usage is metered. "Certain Audio Intelligence features that rely on server-side models are subject to daily usage limits, including but not limited to Live Rewind and Siri Recap," the second footnote states. Apple publishes no numbers for those limits. Its Siri AI materials have said since September that expanded access to server-side features will be available for a fee in the future, with no price or date.

18, not 13

The age floor is where reporting and Apple's documentation part ways. The support page says "Live Rewind and Siri Recap are not available for users under 18," once in its requirements section and again in a footnote. MacRumors reported on November 3 that the features are not available to users under 13, with parental approval required for Child Accounts under 18.

The figure of 13 does appear in Apple's own material, but for a different product. "Siri AI is not available for users under 13," Apple stated in its November 3 developer post opening Siri AI to app developers. Siri AI's minimum age is 13; the two listening features carry a separate minimum of 18. Neither the September press releases nor the overview mentions an age limit, and the support page does not explain how the limit is applied or why it is higher than Siri AI's.

Why this matters for the marketing community

Spoken conversation is becoming a data source on consumer hardware, and two of the largest consumer technology companies building for it have taken different routes. On October 15, Amazon released Bee Frontier, a $74.99 wearable that summarizes conversations. With the $11.99-a-month Bee Pro plan and an Amazon.com connection, products mentioned out loud can surface as recommendations on the Amazon homepage. That is a route from spoken words to a retail media surface, once the wearer opts in.

Apple's documents describe the opposite architecture. Summaries are end-to-end encrypted, Apple says it holds no keys, raw audio is unreachable by apps or the operating system, and the documents describe no route for third-party apps to read Recaps beyond the wearer exporting them. Nothing in Apple's material mentions advertising, personalization or measurement use of Recap data. The monetization path Apple has described for server-side Apple Intelligence runs through usage caps and a future fee for expanded access, not through ads.

The data that does leave the phone still matters to anyone tracking how location context flows. Each Siri Recap request carries the condensed conversation plus labels such as home, work or school, the city, state and country, and place categories such as grocery store or park, according to the overview. Apple's promise is that none of this is stored or visible to it. The reach is limited for now: two watch models on sale since September 18, iPhones from 2024 onward, English only, adults only, and none of it in the EU or China.

For agencies, media sellers and marketing teams, the question sits close to daily work. AI notetakers already join many client calls and pitch meetings, and the lawsuits against Otter.ai and Granola turn on whether everyone in those meetings agreed. A wrist-worn summarizer moves the same question from the calendar invite into the room, with no bot in the participant list and, in Siri Recap's case, no chime.

Timeline

Summary

Who: Apple, through its watchOS 27.3 developer beta, its support page on Audio Intelligence and its September 2026 Privacy Overview. The features affect adult owners of Apple Watch Series 12 and Apple Watch Ultra 4 with a recent iPhone, and the people they talk to.

What: The answer to whether the Apple Watch is listening is yes, once its owner opts in, though Apple says nothing is recorded. Sound Recognition and Shazam listen for sounds and music; Live Rewind keeps a rolling buffer and shows the last 15 seconds of speech as text on request, with a chime; Siri Recap runs an always-on speech detector, sends encrypted audio to the iPhone for transcription and condensing, and sends the condensed text with context such as home or work labels and city to Private Cloud Compute for a summary, with no signal to people nearby.

When: The first watchOS 27.3 developer beta was released on November 3, 2026, and Apple's support page carries the same date. The features were announced on September 9, 2026, and MacRumors expects a public release in December.

Where: English only, outside the EU and China, on Apple Watch Series 12 and Ultra 4 paired with iPhone 16 models (excluding iPhone 16e) or later. Lawyers in California, Massachusetts, Pennsylvania and Washington have questioned how the features fit all-party consent laws.

Why: Apple frames the features as memory aids that let users stay present in conversations, built so that no audio recording exists. The release matters because it puts ambient listening on a mass-market wearable, leaves open whether processing without recording still counts as interception in all-party consent states, and contrasts with Amazon's Bee, which can link spoken items to shopping recommendations.