Skip to main content
CybersecurityPrivacy & Surveillance

Meta's AI Patent Tracks Emotions Through Voice, Biometrics

Person speaking into smartphone in quiet, neutral-colored room with soft daylight.

The application, US 2026/0182881, was filed by Meta Platforms in December 2025 and published on July 2.

Meta's patent application US 2026/0182881: who filed what, when

Meta filed the published application in December 2025; it names a single inventor, Lachlan Dunn, and traces to a provisional filing from December 2024. The patent-analysis site Patentlyze flagged the filing first. The document carries 20 claims: three independent claims cover emotional analysis on its own, while the workout-coach material appears only in dependent claims that build on that emotional-analysis core.

The system's technical scope: continuous listening, multimodal signals, and timestamped logs

The filing describes a device that records spoken audio “across the day.” According to the patent, that device could be any of several form factors:

  • smart glasses
  • a phone
  • a smartwatch
  • headphones
  • a smart home speaker

The recorded speech is transcribed and analyzed by an AI “trained to read mood” that evaluates both the words and paralinguistic cues — tone, pace, sighs, laughs — then tags each stretch of audio with an emotional read. Each read is matched to surrounding context and “time stamped and logged on servers” in at least one figure in the filing. The system builds summaries of patterns over set periods such as a day or a month; one example readout in the filing says, “You sigh most frequently before bed, and you're happiest when with friends. You've expressed more gratitude this month.”

The patent goes beyond voice alone. It allows for biometric and eye-tracking inputs — pupil size, blink rate, even eye moisture — and for signals drawn from device use: the posts a person views or likes, their screen time, and how fast they switch between apps. The filing describes linking individual emotional reads back to the precise words behind them — for example, an anger reading arriving with “the exact harsh words you used.”

Workout coaching as a dependent claim: smart-form feedback and mood-aware coaching

The second strand of the filing is a real-time workout coach. In the patent's examples, smart glasses watch a user’s form in a mirror and speak coaching instructions — telling a user to sink deeper in a squat, cheering on extra reps, or easing off if the system judges the user tired or discouraged. The filing says the coach may “admonish” a user if it decides they are slacking, and claims no human coach could match its precision or sustainment “all day.” The coaching features are presented as building on the emotional-analysis claims rather than as independent inventions.

Amazon Halo and the EU AI Act: precedents and regulatory constraints

The filing sits against two concrete precedents noted in the document. First, Amazon’s Halo product introduced voice mood-reading in 2020 with a Tone feature that listened to pitch and pace and reported emotional impressions; Halo processed samples on-device and deleted them, and the Halo line was shut down in 2023. In December 2020, Senator Amy Klobuchar pressed federal health regulators over Halo’s collection of voice-tone and body-scan data, “calling it unusually intrusive,” according to the filing.

Second, the filing acknowledges regulatory moves in the European Union. Since February 2025 the EU’s AI Act has banned AI that infers people’s emotions in workplaces and schools, with exceptions for medical or safety reasons and penalties up to 35 million euros or 7% of a firm's global turnover. That ban excludes consumer tools; a separate EU rule scheduled to arrive in August 2026 will require systems that read emotions from biometric signals to disclose they are doing so. The filing notes the distinction: a voice-first coach may be arguable under the law, while a system that also reads pupils and blink rate “would fall more squarely inside that biometric line.”

What this means for technologists, policymakers, and end users

  • Technologists and security teams: the claims show emotional analysis as the primary capability, with both on-device and server-logged variants; engineers and privacy teams will face concrete design choices about where analysis runs and whether logs are kept “time stamped and logged on servers.”
  • Policymakers and regulators: regulators have an active legal framework to apply — the EU AI Act’s workplace-and-school ban and the August 2026 biometric-disclosure rule are explicit touchpoints the filing itself cites.
  • End users and the general public: the filing describes a persistent, timestamped log of emotional reads keyed to location, activity, and device behavior — a continuity of records that the document contrasts with earlier efforts such as Amazon Halo and that, for now, exists only on paper because Meta has not announced any product based on the filing.

The filing stakes a claim on an ambitious combination of continuous listening, biometric sensing, and behavioral telemetry. As the document itself notes, some versions place processing on-device while others log data to servers, and none of the described systems ships in a product today. The Hacker News has reached out to Meta for comment on whether the application reflects product plans or how user data would be handled; that outreach is noted in the filing’s reporting and any response would change the public record. For now, the patent places an idea on the table — and the filing's final, blunt line captures the immediate reality: Amazon’s earlier, narrower attempt was pulled; Meta’s version reaches farther, and “the only thing keeping it out of your life is that no one has built it yet.”

https://thehackernews.com/2026/07/meta-files-patent-for-ai-that-can.html