China's government, which runs an estimated 700-million-camera CCTV network, has published a privacy code specifically for smart glasses: wearable cameras worn on the face that can record bystanders without their knowledge. The voluntary code, issued by a research institute under the Ministry of Industry and Information Technology, asks manufacturers to get explicit consent before recording, equip devices with hardware LED indicators that show when a camera is active, and collect as little data as possible (the Gizmodo report; Sina Finance confirms the document is titled the "AI Glasses Trusted Vision Self-Discipline Covenant" (自律公约), framed as voluntary industry guidance rather than statute).
The three concrete asks matter because they name specific failure modes any hardware maker could fix tomorrow. An LED indicator that anyone in the room can see is already present on most smart glasses sold in China and the United States, but no country currently mandates one. Consent before the shutter trips is what a recent Rokid incident exposed as missing. Minimum data collection keeps the raw footage on the device instead of sending every frame to a server. CNR's tech desk frames the move as a "dual tech-plus-regulation" approach to covert recording (CNR); the South China Morning Post places it inside Beijing's broader priority of treating AI eyewear as a privacy-sensitive product category (SCMP).
The trigger was a small, ugly incident. Earlier this year, Rokid smart glasses recorded Spring Airlines flight attendants during boarding; the videos were posted to Rokid's own community platform. Within days, third-party stickers marketed to obscure the LED indicator surfaced online, a proof-of-concept that any hardware safeguard can be defeated by a $5 sticker.
That fragility is not a China-only problem. In the United States, Meta sells the Ray-Ban Meta smart glasses with a privacy-controlled design and a small recording LED. Gizmodo's Joanna Stern has documented users drilling out the LED for cash, and a Svenska Dagbladet and Göteborgs-Posten investigation found that Sama, a Meta subcontractor in Nairobi, had workers reviewing Meta Ray-Ban AI training videos that captured bathroom visits, sex, and other intimate moments recorded by the glasses (Svenska Dagbladet). The consent-by-indicator model is breaking even where the hardware exists.
The asymmetry for a U.S. reader is sharp. China's voluntary national code asks for three specific things. The United States has no federal LED requirement, no federal consent rule, and only a state-by-state patchwork of general recording laws that predate the smart-glasses category. China is at least writing down what it expects. The U.S. is letting the gap widen by default.
The voluntary caveat is the whole point. 自律公约, literally "self-discipline covenant," has no enforcement mechanism. Manufacturers sign on; manufacturers can walk away. Rokid, whose glasses started this round of attention, has not publicly committed to the covenant. The same government that wrote the code operates the world's largest public camera network. The document is a statement of what the regulator thinks the problem looks like, not a fix.
What other regulators, platforms, and hardware makers do with the document is the more interesting question. The three asks, explicit consent, hardware indicator, minimum data, are portable enough that any of them could be lifted into a binding rule by California, Brussels, or a platform's own developer policy. The first company to ship a smart-glasses product with a tamper-resistant LED and a default-off recording mode will own the privacy position Meta and Rokid currently cannot. China's voluntary code is not the answer. It is the cleanest current list of the questions.