Meta's smart glasses companion app was downloaded more than 50 million times before anyone disclosed that it already contained three AI models capable of detecting a face, generating a biometric finge
A security researcher, Buchodi, said Meta’s Stella companion app for Ray-Ban and Oakley smart glasses (Android build 273.0.0.21) contained a dormant on-device facial recognition pipeline, including three AI models (SCRFD, KPSAligner, SFace), a SQLite “person_profiles” schema, and a vector index for 2048-dimension biometric embeddings. WIRED reported the code was added since January 2026, and the app had been downloaded 50M+ times before disclosure. Meta said nothing shipped to consumers and no f
Raises near-term regulatory, litigation, and reputational risk around Meta’s smart-glasses biometric plans, even if the feature is not enabled for typical users.
Meta’s Ray-Ban/Oakley companion app reportedly contains dormant on-device facial recognition models and a biometric database schema shipped to 50M+ downloads.
Moderate downside bias on headlines; material impact depends on regulator/lawmaker follow-through and any confirmation of consumer-facing activation.
Background
Meta has faced prior biometric backlash: deletion of 1B+ Facebook faceprints (2021) and large settlements (Illinois $650M; Texas $1.4B). Recent reporting (NYT Feb 2026) described revived smart-glasses facial recognition plans under an internal “NameTag” concept.
Why it matters
The Buchodi/WIRED code review adds technical specificity: three on-device models (face detection, alignment, and 2048-dimension biometric embedding) plus a person_profiles schema and a write path for unrecognized faces, shipped in app updates to tens of millions of downloads—though Meta says it is not enabled for consumers.
Market relevance
This is a privacy/regulatory catalyst: shipped dormant biometric capability can trigger investigations, compliance changes, and litigation risk even without confirmed consumer activation.
Market effects
Biometric/AI privacy scrutiny risk increases for consumer AR/wearables and on-device AI deployments; could pressure peers’ product roadmaps and compliance spend.
US-focused risk given Illinois/Texas biometric law history and potential new regulator interpretations of on-device faceprint collection.
May spill into broader EU/UK privacy enforcement narratives around consent, data minimization, and transparency for biometric processing.
Alternative perspectives
Because the recognition pipeline appears inactive for unenrolled accounts and Meta frames this as “evidence” of development, the market may overreact versus the actual probability of consumer-facing deployment.
Regulators may focus on whether the app’s write path and RLDrive server-side updates constitute “collection” under biometric statutes; confirmation of any data transmission would be the key swing factor.
Key entities
- companyMeta
Developer of Ray-Ban and Oakley smart glasses and the Stella companion app containing dormant facial recognition infrastructure per WIRED/Buchodi.
- softwareStella (companion app)
Android companion app for Meta’s smart glasses; analyzed in version 273.0.0.21 for facial recognition pipeline components.
- softwareRLDrive
Meta’s cross-device sync framework referenced as the mechanism for server-side updates to the face namespace.
- personSen. Ed Markey
Lawmaker who reportedly wrote to Meta in March 2026 seeking information on facial recognition plans.
