--- title: "Who Just Recognized You on the Street? Smart Glasses and the End of Passive Anonymity. The Take Inside China" date: 2026-08-30 originalDate: 2026-08-18 originalTitle: "AI眼镜来了，谁认出了你？街头人脸识别的边界之争" issue: "Issue 4" description: "Meta's dormant NameTag code, a Pentagon-supplier recognition license, Florida sheriffs already wearing glasses — read through Chinese eyes." tags: - "smart glasses" - "facial recognition" - "Meta" - "privacy" - "surveillance" sources: - label: "WIRED: Meta silently added face-recognition code for its smart glasses (NameTag, June 2026)" url: "https://www.wired.com/story/meta-smart-glasses-face-recognition-nametag-connections/" - label: "WIRED: Meta tapped a Pentagon supplier to prototype face recognition (Rank One Computing)" url: "https://www.wired.com/story/meta-rank-one-computing-face-recognition-smart-glasses/" - label: "EFF: Victory — Meta strips facial-recognition code from smart-glasses app after public outcry" url: "https://www.eff.org/deeplinks/2026/06/victory-meta-strips-facial-recognition-code-smart-glasses-app-after-public-outcry" faq: - q: "What was NameTag?" a: "Per WIRED's June 2026 code review: a dormant, unreleased facial-recognition system inside the Meta AI companion app that powers Ray-Ban and Oakley smart glasses (installed on 50M+ phones). Design: glasses see a face → match against biometric faceprints pulled from Meta's servers and stored on the user's device → alert the wearer who it is. The code even contained a social hook — recognized people could be sent friend requests." - q: "What happened after the discovery?" a: "Meta shipped an update removing the code the day after WIRED's report; the EFF called the removal a victory after public outcry. But two context facts survive the deletion: the recognition tech was licensed from Rank One Computing, a Denver firm whose customers include the US Marshals Service and police agencies and whose board includes a former CIA deputy director and former FBI science chief; and a newly public Meta patent describes recognition results feeding automatically generated photo collections. Meta's position: the feature was never enabled and user control is paramount." - q: "Are police already wearing these?" a: "In Florida, yes. Per records Miami New Times obtained under the state's Sunshine Law, officers in at least two sheriff's offices have worn smart glasses on street patrol; Gizmodo's headline: smart glasses are 'catching on' with US police. Meanwhile the consumer side moved the other way — Kylie Jenner's AI-glasses endorsement triggered a fresh privacy backlash, 70+ civil-rights organizations signed a joint letter against adding facial recognition, and some San Francisco bars ban Meta glasses at the door." - q: "Why does a Chinese column care about US glasses?" a: "Three layers, per the take. Hardware is global: Chinese AI-glasses makers ship camera+AI as standard, and a validated feature path (recognize → label → social/album) gets cloned subject only to time and compliance. China has actually been braking on face-scraping — Beijing, Shanghai and Shenzhen hotels have dropped forced face check-in under separate-consent-plus-necessity rules — but those rules assume a visible camera. A passerby's glasses are zero-notice, zero-consent, zero-record. And the essential shift: privacy fights used to be about whether institutions may identify you; glasses hand identification to every random pedestrian." --- Have you made eye contact with a stranger wearing glasses today? If so, the opening question of our WeChat column's piece (2026-08-18, "AI glasses are here — who just recognized you?") applies: during that one second, did they know who you are? The occasion was WIRED's June 2026 discovery of dormant face-recognition code inside Meta's smart-glasses app, plus the fact that Florida sheriff's deputies were already wearing similar glasses on patrol. This entry translates the take, pinned to the WIRED/EFF record. ## The numbers - NameTag's host app: installed on 50M+ phones (WIRED) - Timeline: WIRED report → Meta removes code the next day (June 2026); EFF logs it as a campaign victory - Police adoption: at least 2 Florida sheriff's offices, street patrol (Miami New Times via Sunshine-Law records, as cited) - Consumer resistance: 70+ organizations in a joint letter (April, as cited); San Francisco venues banning the glasses at the door - **Figure arbitration:** none needed — the load-bearing facts are dated and sourced; the take's "millions of pairs sold" for hardware base rides on the 50M+ app-install figure. ## The take inside China The essay's method: resist the easy "Meta repented" reading by stacking three lines of evidence — the code existed, the patent was filed, and the supplier comes from the military-police sector. Deletion, it argues, is a pause button, not a delete key: with hardware already on millions of faces, re-enabling is one OTA update away. Then the diverging-votes image that anchors the piece: **consumers vote with their feet** (endorsement backlash, bar bans) while **police departments vote with their budgets** (procurement records don't lie) — the two streams flowing opposite directions is itself the most truthful caption on the story. The third section is the one that makes this a China Watch entry rather than a US-tech note. China's regulators have spent the year *tightening* facial recognition — hotels in major cities dropping forced face check-in, the separate-consent-and-necessity principle repeatedly affirmed. The take's point: all of that presumes a visible, institutional camera. Glasses abolish the premise — collection with zero notice, zero consent, zero record, performed by random passersby for the price of a wearable. The reframe: past privacy fights asked whether *institutions* could identify you; this is the first technology that degrades the anonymity of the street itself, the who-knows-nobody city. When identification costs the price of a pair of glasses, anonymity becomes a luxury someone pays for. Fairness, as the take insists on delivering: the positive list is real — finding missing persons, real-time captions for the hearing-impaired, instant translation for travelers. The hard line to draw is between banning it and abusing it. ## What the Chinese take left out Jurisdictional follow-through: what EU AI Act biometric-categorization rules would do to a glasses feature in Europe, and any MPIP/regulatory comment on passively collected face data — both absent. Also no US statutory picture beyond the org-letter; the Senate/state-level facial-recognition bills that an English reader would expect in a piece about America. ## Why it matters outside This is the wearable end of a general curve: recognition cost collapsing from datacenter to consumer device. The China reading adds the sharpest formulation we've seen — anonymity shifting from default to paid luxury — and the diverging-votes frame (consumer backlash vs police procurement) is the two-line summary of how facial recognition actually diffuses. For policy readers: China's separate-consent regime and the US's sectoral patchwork are both about to meet the same zero-notice collector neither was written for. ## Sources - [WIRED: Meta silently added face-recognition code for its smart glasses](https://www.wired.com/story/meta-smart-glasses-face-recognition-nametag-connections/) - [WIRED: Meta tapped a Pentagon supplier (Rank One Computing)](https://www.wired.com/story/meta-rank-one-computing-face-recognition-smart-glasses/) - [EFF: Meta strips facial-recognition code after public outcry](https://www.eff.org/deeplinks/2026/06/victory-meta-strips-facial-recognition-code-smart-glasses-app-after-public-outcry) > **Provenance & disclosure.** Originally published in Chinese on our WeChat channel on 2026-08-18 ("AI眼镜来了，谁认出了你？街头人脸识别的边界之争"); drafted with AI assistance under human editorial direction. Translated to English on 2026-08-30 (AI-assisted, human-reviewed). The NameTag chronology, Rank One details and removal were checked against WIRED's two reports and the EFF post; the Florida-patrol count, org-letter count and hotel-face-checkin rollbacks remain as-cited [unverified]. This is translated commentary — not a SigPulse measurement. Our first-party measurements live in the [dispatches](/posts/) and the [/data/ ledger](/data/).