Meta Patents AI System That Listens to Users' Moods All Day

The US Patent and Trademark Office published a patent application from Meta on July 2, 2026, describing an artificial intelligence system designed to listen to a user's voice and ambient sounds throughout the day in order to draw conclusions about their current mood. The document, first spotted by Patentlyze, describes a device that Meta itself compares to a personal trainer.

How the system works

According to the patent description, the device records the user's speech continuously, without requiring a wake word or a separate recording session to be started. The recorded audio is transcribed into text and then run through a machine learning model that evaluates both the content of what is said and how it is said.

The patent document states outright that voice communications are to be transcribed, and that the machine learning model assessing emotional state is meant to interpret verbal and non-verbal cues to determine emotional indicators. On top of that comes situational context: time of day, location, the user's activity, and digital interactions, meaning what content they browse, how long they use the screen, and which apps they switch between.

The patent also describes the use of biometric signals such as pupil size, blink rate, and eye moisture, which combined with voice analysis are meant to give a fuller picture of the user's wellbeing at any given moment.

Stated purpose versus real risk

Meta presents the whole concept as a fitness tool. The system is meant to select exercises and correct training technique depending on whether the user is tired, stressed, or energized at a given moment. The patent description includes an example summary a user might receive: a note that they sigh most often before falling asleep, and that they tend to be happiest in the company of friends.

404 Media points out, however, that a device listening all day inevitably also records conversations of third parties who never consented to it, along with very detailed data about the user's location. The patent does not address this issue. Critics also note that Meta's business model relies on advertising tailored to users' interests and behavior, which raises the question of whether emotional data would actually be used solely for training suggestions.

Voice communications may be transcribed, and the machine learning model assessing emotional state may interpret verbal and non-verbal cues to determine emotional indicators - from Meta's patent description

Meta's response

Meta spokesperson Tracy Clayton addressed the matter with a caveat typical of such situations, noting that patent applications are often used to lay out ideas that may or may not ever be implemented. This is a standard response from tech companies to reports about controversial patents, which don't automatically mean the described device will reach the market.

What it means for Poland and the EU

If Meta were to actually deploy such a system in the European Union, it would have to contend with two parallel legal regimes. The first is the GDPR, which requires clear, informed consent for data processing, and continuous recording of the surrounding environment without an explicit trigger is hard to reconcile with the principles of data minimization and purpose limitation.

The second is the EU AI Act, which classifies emotion recognition systems as high-risk and, since February 2025, explicitly bans their use in workplaces and educational settings. In addition, starting in August 2026, new rules take effect requiring users to be informed when biometric emotion recognition systems are in use, which for the device described in Meta's patent would mean far more transparent communication than what appears in the filing itself.

For now, the patent remains a legal document protecting intellectual property, not an announcement of a specific product. The history of similar projects, like the Amazon Halo wristband discontinued after three years, shows however that even companies with substantial resources back away from continuous voice listening when they run into pushback from the public and regulators.