Meta has secured a patent for an AI system that continuously monitors your voice and gauges your mood throughout the day.

      Meta has secured a patent for an AI system that persistently records users' voices, transcribes audio, and analyzes it using a machine learning model to ascertain their emotional state. The patent, published on July 2, outlines a device that listens for various "audible communications," such as sighs, laughter, and vocal inflection, while integrating “contextual factors” like time of day, location, activities, and medication schedules. This system creates a continuous mood log derived from these inputs.

      Identified by Patentlyze, the patent positions the technology as a fitness device. Meta contends that AI-driven emotional coaching could enhance workout techniques and offer guidance that a human trainer might not provide. However, the described architecture extends beyond just exercise sessions, aiming for "continuous emotional monitoring on everyday devices" through what the patent refers to as “multimodal sensor inputs on synchronized timelines.” It captures tone, pace, pauses, and breathing at “predefined times,” quantifying the user's emotional state and enabling correlations between mood and medication timings, yielding insights like “a more positive emotional state linked to specific times of day or around medication intake.”

      Amazon attempted a similar initiative with its Halo Band in 2020, which featured a microphone for “tone of voice analysis.” However, following public backlash, Amazon removed the microphones in the 2021 model and ultimately ceased the product line in 2023. Meanwhile, Meta’s smart glasses are already embroiled in a privacy controversy over secret recording, with seven million units sold and two lawsuits in the US claiming the company misled consumers regarding footage usage. Incorporating continuous ambient voice recording into their product line would further expand surveillance from visual to emotional monitoring.

      A Meta representative informed 404 Media that “patents at Meta are often filed to disclose concepts that may or may not be implemented,” which is typical corporate caution. However, the patent’s detailed nature, including medication tracking and emotion-time correlations, implies more than just a speculative filing. In Kenya, data workers already faced job losses after it was revealed they assessed sensitive footage recorded by Meta’s glasses. Should this patent be developed into a product, there will be a need for someone to train the emotional model, which would involve listening to recordings of strangers expressing emotions at home and assigning mood scores for the algorithm to learn to identify feelings like sadness.