Get the newsletter
Meta · Chip · Filed Feb 10, 2026 · USPTO published Aug 20, 2026 · Aug 20 release

Meta Patents a Dedicated Security Chip for Biometric Data in Wearables

Meta is designing smart glasses and wristbands that store your biometric data on a completely separate, locked-down chip that the rest of the device's software simply cannot reach.

Smart glasses paired with a dedicated security chip architecture for biometric data processing. Drawing from patent filing US 2026/0244726 A1.
Smart glasses paired with a dedicated security chip architecture for biometric data processing.All 9 drawings ↓
Publication number
US 2026/0244726 A1(US20260244726A1)
Applicant
Meta Platforms Technologies, LLC
Filed · Published
Feb 10, 2026 · Aug 20, 2026
Inventors
Jason Haensly, Hsuhua Yu, Bo-Huei Lin, Anish Kumar Goyal, Avdhesh Chhodavdia
Status when we published
Waiting for an examiner
Document
20 claims
US classification
726/22
Parent application
Claims priority from a provisional application 63760099 (filed Feb 18, 2025)

What Meta's isolated biometric chip actually does

Imagine unlocking your smart glasses with your fingerprint, and then wondering: where does that fingerprint data actually live, and who can see it? That's the exact problem Meta is trying to solve.

Meta's idea is to add a second, smaller chip to its wearable devices alongside the main processor. This companion chip handles all biometric data, like fingerprints or face scans, entirely on its own. The main chip, which runs your apps and connects to the internet, is physically blocked from ever reading that data. It's a bit like keeping your safe inside a room with a separate lock that your housemates don't have a key for.

The companion chip uses a small AI model to check whether incoming sensor data looks legitimate (not a fake or a replay attack), then compares it against encrypted templates stored in a protected vault. The main processor only ever learns one thing: yes or no, authentication passed.

From the filing · CLAIM 1
a machine-learning (ML) accelerator for biometric processing, wherein the ML accelerator is configured to: generate validated sensor data by applying a biometric sensor data to an ML model, and generate a biometric signature from the validated sensor data; …

Translation: A specialized mini processor handles the heavy lifting of turning raw physical scans into secure digital login keys.

How the companion SoC walls off your biometric data

The companion SoC (system-on-a-chip, meaning a self-contained processor) is purpose-built to be small and low-power, important constraints for wearables like glasses or wristbands. It contains two key parts that never directly talk to each other except through a tightly controlled internal bus.

The first part is an ML accelerator, a dedicated processor for running AI models. It takes raw sensor data (say, a fingerprint scan) and passes it through a machine-learning model that acts as a gatekeeper: it checks whether the data is real and valid before doing anything further. If it passes, the accelerator generates a biometric signature, a compact mathematical representation of the scan.

The second part is an isolated secure subsystem. It stores your encrypted biometric templates (the reference scans enrolled during setup) and compares incoming signatures against them. Crucially, it also encrypts the AI models themselves and verifies their authenticity before letting the accelerator use them, so an attacker cannot swap in a tampered model to fool the system.

  • A network-on-chip (NoC) connects the two parts internally
  • Hardware firewalls in that network ensure raw sensor data can only go to the ML accelerator
  • Stored templates can only be read by the secure subsystem

The main SoC, which handles everything else the device does, is locked out entirely.

From the filing · THE ABSTRACT
… a subsystem, isolated from general purpose processing of the main SoC, interconnected through a network-on-chip with access control firewalls restricting access to biometric sensor data to the ML accelerator and stored biometric templates to the subsystem.

Translation: A heavily guarded internal vault keeps your sensitive body data completely separated from the watch's normal operating system.

What this means for privacy in Meta's wearables

Biometric data is permanent. You can change a password; you cannot change your fingerprints. So if a wearable device leaks that data through a software bug or a compromised app, the damage is irreversible. Meta's approach moves the most sensitive data onto a chip that is physically isolated from the software layer where most attacks happen, meaning a hacked app or even a compromised operating system on the main processor cannot reach your biometric templates.

For anyone wearing a Meta device that uses facial recognition or fingerprint login, the practical payoff is that your biometric data stays locked in hardware you do not have to think about. The architecture also verifies the AI models themselves before use, which closes a subtle attack path where an adversary replaces the recognition model with one designed to accept fakes. This kind of chip-level security work is part of a broader wave of new Big Tech patents focused on hardening biometric systems inside wearable and AR hardware.

Meta's 51st filing in our smart glasses watchlist since May adds to a picture that includes circuits inside clear lenses and gesture and gaze learning.

Source: US 2026/0244726 A1 at the USPTO (PDF). Full patent text and figures are in the official publication.

Get our take in your Top Stories

Liked this breakdown? Add Patentlyze as a preferred source on Google, and our plain-English take shows up more often in your Top Stories the next time Meta patent news breaks.

There are more where this came from

We read every patent application Big Tech publishes and send you the ones worth knowing. Plain English, free, every week.

The drawings · 9 sheets

9 drawing sheets from US 2026/0244726 A1 · click any drawing to enlarge

Smart glasses security architecture diagram, drawing sheet 1 from US 2026/0244726 A1
Sheet 1 Smart glasses security architecture diagram
Co-processor biometric authentication layout, drawing sheet 2 from US 2026/0244726 A1
Sheet 2 Co-processor biometric authentication layout
Enrollment request processing workflow, drawing sheet 3 from US 2026/0244726 A1
Sheet 3 Enrollment request processing workflow
Machine learning model verification sequence, drawing sheet 4 from US 2026/0244726 A1
Sheet 4 Machine learning model verification sequence
Authentication attempt processing steps, drawing sheet 5 from US 2026/0244726 A1
Sheet 5 Authentication attempt processing steps
New user template enrollment process, drawing sheet 6 from US 2026/0244726 A1
Sheet 6 New user template enrollment process
Extended reality meeting environment overview, drawing sheet 7 from US 2026/0244726 A1
Sheet 7 Extended reality meeting environment overview
Message typing in extended reality, drawing sheet 8 from US 2026/0244726 A1
Sheet 8 Message typing in extended reality
Show the other 1 sheets
Meta Patents a Dedicated Security Chip for Biometric Data in Wearables patent drawing, sheet 9
Sheet 9
Patent filing page

Watchlists this filing belongs to

All watchlists →