Kaspersky raises security alarms over Apple Watch Series 12

Apple launched the Apple Watch Series 12 and Apple Watch Ultra 4 in September 2026 across global retail outlets with ambient listening tools. The integrated Audio Intelligence system deploys the built-in microphone to track surrounding sounds, recognize songs, and log spoken dialogue for automated note generation.
Kaspersky lead security researcher Fabio Marenghi examined the privacy implications of persistent acoustic monitoring alongside ESET Latinoamérica investigator Mario Micucci. Both analysts identified specific risks tied to portable consumer hardware processing ambient speech without manual activation. Their technical assessment focuses on the operational surface exposed by automated recording routines.
Hardware integration powers active listening in S11 architecture
The Audio Intelligence software layer operates directly on the S11 system-in-package inside both the Series 12 and Ultra 4 smartwatches. This custom silicon manages mechanical audio reception to capture sounds that users might overlook during typical daily activities.
The audio framework bundles Sound Recognition, Shazam identification, Live Rewind, and Siri Recap into a single processing pipeline. Sound Recognition generates haptic pulses upon detecting doorbells, sirens, vehicle horns, or crying infants to assist hearing-impaired users. Shazam recognizes background music tracks independently without relying on a paired iPhone connection.
Live Rewind captures ongoing conversations through an interface shortcut that stores the preceding 15 seconds of spoken interaction. Users can convert that buffer into transcriptions saved directly to the Siri storage bank.
Siri Recap applies machine learning to daily acoustic recordings to produce structured outlines rather than verbatim audio files. Apple asserted that this synthesis relieves consumers from the manual chore of logging verbal appointments during meetings. The utility focuses on organizing calendar commitments and spoken reminders.
Daily briefing mechanics structure automated voice logs
The Siri framework converts ambient conversational cues into categorical notes before assembling a consolidated daily agenda. Apple stated that the software architecture restricts output to structured summaries rather than full audio transcripts.
Owners manage specific operating intervals through internal Apple Watch settings. The interface allows wearers to halt continuous acoustic monitoring during night hours or block logging entirely within office buildings. These scheduled parameters give users manual control over microphone availability.
Technical safeguards confront forensic data extraction methods
Apple documented its user protections through on-device computation, file encryption, and cryptographic identity masking that prevents the archiving of raw audio on external infrastructure. Corporate specifications indicate that system algorithms do not associate captured voices with verified user identities.
The S11 silicon routes audio streams through an isolated hardware enclave that operates separately from standard device storage, purging source buffers immediately after transcription runs. The firmware also sanitizes banking credentials and sensitive financial numbers before generating written summaries. These software filters prevent payment figures from entering final notes.
Live Rewind sounds an audible alert tone whenever transcription routines activate. That mechanical cue sounds even if the user selects silent mode on the wrist interface. The watch screen simultaneously shows a bright microphone emblem to notify nearby individuals about the active capture window.
External cybersecurity specialists maintain reservations regarding ambient microphone operations despite these manufacturer safeguards. Fabio Marenghi stated that secure implementation requires shared operational discipline between device makers and end users. He emphasized that local storage protections must remain resilient against credential misuse.

“The manufacturer must guarantee that this audio processing takes place locally rather than on cloud infrastructure where files face greater leakage hazards,” said Marenghi, noting that primary account holders must also maintain strong credentials to prevent external takeover.
Consumer privacy debates expand because the Apple Watch chassis positions the intake sensor outward to sample atmospheric sound throughout the day. Technical teams continue testing the operating system to catalog what sensitive data points remain vulnerable if malicious actors breach the perimeter. Unresolved vulnerabilities could expose surrounding workplace communications.
Hardware safeguards isolate audio data.
The legal settlement required Apple to disburse $95 million in 2025 to resolve a class-action lawsuit in the United States alleging that unauthorized Siri recordings captured private household conversations and forwarded relevant contextual metrics to targeted advertising platforms. Apple denied all wrongdoing throughout the courtroom proceedings before settling the claim.
Mario Micucci warned that technical exposure extends beyond raw sound files because stored written summaries reveal personal routines if compromised. The ESET investigator evaluated how local databases interact with backup directories across synchronized platforms. His analysis focused on cross-app data routing.
“Even without archiving raw sound, an organized text recap can expose sensitive intelligence,” said Micucci. “Apple notes that stored texts sync through cryptographic pipelines, yet access to an unlocked handset or data transfers toward external programs generate attack avenues.”
Exfiltrated briefing notes present severe exposures regarding corporate espionage, identity extortion, and unauthorized habit tracking by online criminal groups. Intruders target daily calendars and task logs to identify business targets. Intercepted summaries allow attackers to map personal vulnerabilities.
“With contemporary AI progress, synthetic voice duplication stands among the most dangerous developments,” warned Marenghi. He noted that threat actors harvest short vocal samples to impersonate executives and execute fraudulent payment scams against relatives through WhatsApp or live phone calls.
Passive recording risks affect non-users who speak within listening range of the smartwatch without knowing about the active microphone. Wearable sensors collect statements from coworkers and passersby who never agreed to corporate terms. In Brazil, statutory regulators evaluate these passive practices under the provisions of the LGPD data protection statute.
“The LGPD protects personal data whenever information relates to an identifiable individual, even if algorithms execute tasks without keeping raw recordings,” Micucci said. He stated that individual consent given by the device buyer does not extend legal authorization to collect audio from third parties.
Wearable audio alternatives operate across retail electronics
Autonomous transcription and conversational summarization extend across multiple consumer hardware brands beyond Apple. Competitive device manufacturers market wearable tools equipped with machine learning recorders. Competing firms offer alternative hardware platforms.
The Plaud Notepin utilizes artificial intelligence engines to record spoken dialogue and assemble executive meeting transcripts. The lightweight wearable accessory fastens to clothing and functions quietly while processing verbal discussions into formatted text.
E-commerce retailers such as Amazon distribute discreet pocket pens and miniature voice recorders containing internal flash memory units. These units capture local audio environments for subsequent manual export. Specialized vendors provide distinct recording solutions for varied consumers.
Contemporary smartphones and previous Apple Watch generations execute comparable transcription tasks by pairing with third-party software tools. Applications like Wave process live cellular calls to generate immediate written transcripts and factual summaries directly on mobile displays.






