The Paper tests multiple AI glasses: a light-blocking sticker costing under 10 yuan can bypass the recording indicator light

Main body of the article

This sticker is called a "light-blocking sticker" and sells for less than 10 yuan.

Recently, The Paper reporters tested Huawei, Xiaomi, Rokid, Qwen, RayNeo, INMO, Meizu and other mainstream AI glasses on the market and found that for most mainstream models, even after the indicator light is completely covered by a light-blocking sticker, they can still shoot normally and will not trigger a blocking alarm; only a few models cannot continue shooting.

It is worth noting that the related topic attracted attention earlier this year. Previously, the Rokid glasses community was reported for a community user secretly filming flight attendants, which sparked heated public discussion. Rokid subsequently deleted the related video and upgraded the glasses' indicator light detection mechanism.

A Rokid glasses community user posted a secretly taken photo of a flight attendant

However, according to a merchant on the platform who spoke to reporters, although the first-generation light-blocking sticker became ineffective as a result, the product has already been updated to a second generation that "can handle anything."

On multiple e-commerce platforms, such light-blocking stickers are sold openly, with sales on one platform exceeding 5,000 units; on Xianyu, a seller directly states on the product detail page that "secret photography will not lead to discovery"; on one e-commerce platform, a merchant told reporters that many buyers purchase light-blocking stickers so that "secret photography is less conspicuous and awkward," and long-term wholesale customers can enjoy discounts.

Various light-blocking sticker products displayed on an e-commerce platform

The related risks are spreading: with the rise of emerging AI technologies, cheating and covert recording are occurring frequently. How should privacy recording be controlled? Where does personal information data go? According to a college entrance examination warning issued by the Ministry of Education on June 2, bringing smart glasses into an examination room is deemed cheating regardless of whether shooting is actually enabled; South China Agricultural University reported in July this year multiple cases of smart glasses being illegally brought into examination rooms in attempted cheating.

On Xiaohongshu, some bloggers also posted about this hidden danger, and someone in the comments replied: "You're so ugly, you won't be secretly photographed, don't worry." The discussion about privacy then slid into evaluations of appearance.

"Suitable for taking photos and recording videos when you don't want to get beaten"

Similar smart glasses light-blocking stickers are openly sold on multiple e-commerce platforms. On Xianyu, for example, a seller explicitly states they are "suitable for when you don't want to be seen taking photos or recording videos, and don't want others to beat you up."

The Paper's investigation found that on an e-commerce platform, both the product page screenshot and the product copy for light-blocking stickers point to covert shooting needs

Specifically, AI glasses usually have a flashing indicator light while shooting, so it can be observed that the AI glasses are in shooting mode. The purchased black round sticker can be placed over the glasses' indicator light (usually on the left side). After covering it, when the AI glasses are used to shoot, the originally flashing indicator light completely turns black, and the shooting state cannot be observed with the naked eye, but shooting can proceed normally.

The Paper reporter tested on site: after applying the light-blocking sticker, the indicator light went out

"Currently AI glasses are all compatible. Basically, after they take mine back, I haven't seen any problems." The merchant told reporters, frankly saying that many buyers purchase it for "secret photography that is less conspicuous and awkward." He also said he has many long-term VIP customers, and as long as the purchase quantity is large enough, he can offer a certain discount.

Why does "one sticker defeat it"?

On the platform, a buyer review of a smart glasses light-blocking sticker mentioned: at first it will alarm about blocking, but after adjusting the light-leak hole to the right position, it will no longer alarm. In addition, according to the alarm characteristics of different brands, merchants also customize "exclusive versions."

In response, the above merchant told reporters that the platform will not introduce relevant regulation, because "we haven't done anything illegal; some people are just being particular."

Why does the phenomenon of "one sticker defeats it" occur? Zhong Junhao, secretary-general of the Shanghai Artificial Intelligence Industry Association, told The Paper reporters that the privacy notification mechanisms of most existing smart glasses are largely ceremonial notices, and indeed have the hard flaw of being defeated by a single sticker.

For example: the shooting notification on many consumer models relies on an independent indicator light and software-layer occlusion detection, and the two are often not fully bound to the imaging chain in engineering terms; light-blocking stickers mainly change the visible-light path or photosensitive readings, and the system sometimes finds it difficult to distinguish whether it is sticker occlusion, lens fogging, or normal environmental dimming such as entering a tunnel, so the shooting process may still continue.

"This privacy notification is like a fuse that can be cut by human action; software-level detection is easily bypassed by physical means. In industry research, we have noticed that on some models the notification sound can only be heard by the wearer, the indicator light is hidden inside the temple, and the light flashes only for an instant when taking a photo, making the shooting behavior very difficult to notice from outside," Zhong Junhao mentioned.

According to foreign media reports, Meta's previously launched smart glasses were dubbed "pervert glasses" because they were used to secretly film others undressing. Some users filmed more than 350 videos of approaching women with Meta smart glasses, and some of those filmed were even doxxed.

More than 70 users who bought, used, or were filmed by Meta smart glasses have sued Meta, claiming that their private and sensitive photos were leaked to overseas workers hired by the company. In addition, according to foreign media reports, relevant governments have planned to temporarily ban the use of glasses with cameras in public places such as parks, schools, and beaches, and said they will set up an expert group to investigate the challenges brought by wearable technology and provide recommendations on permanent regulation.

Companies call on peers to respond collectively and actively

Zhong Junhao explicitly told reporters that for AI glasses to move toward all-day perception, a core boundary must first be drawn: the glasses hardware can "see all day," but it cannot "record all day and upload all day" without restraint. A wearer can wear the glasses to perceive the environment around the clock, but if after a day of wearing, sensitive data such as environmental images, faces, and voiceprints are continuously retained and controlled by a company's backend, that has already crossed the privacy red line.

A number of companies also responded to reporters regarding privacy-related controversies. AI glasses company RayNeo told The Paper that the development of smart glasses must be built on full respect for the privacy and security of users and the public. RayNeo has now integrated its privacy protection system throughout the entire lifecycle of product design, research and development, production, and use. In response to the shooting privacy issues that smart glasses may involve, RayNeo has established multiple protective mechanisms at the product hardware design level.

"We do not support or encourage such behavior. In the future, we will continue to invest in privacy and security technology research and development, strengthen product safety design and user education, and promote the healthy and orderly development of the smart glasses industry," RayNeo said.

"Privacy protection is a relationship between spear and shield, something that must always be done. Companies can only increase the cost of making mistakes as much as possible." Zhu Mingming, CEO of another leading AI glasses company, Rokid, told reporters that in response to a vulnerability in the previous generation of products whereby infrared could be used to bypass the shooting indicator light protection, Rokid has completed the corresponding technical fix. However, he also admitted frankly that relying solely on technical means cannot completely eliminate privacy and security risks.

Zhu Mingming introduced that the Japanese market was highly sensitive to glasses cameras in the early days, but now it has accepted Rokid-related products; European policymakers remain relatively cautious about such devices, but there is still considerable local user demand for purchasing them. In his view, privacy risks require collaborative governance by multiple parties. Beyond technical defenses, user self-discipline, social rules, punishment mechanisms for violations, and guidance at the regulatory level are equally crucial.

Zhu Mingming called on the industry, hoping that more practitioners will jointly confront this public issue. "We may not immediately find a perfect solution, but I hope more peers will actively take a stance and work together to address privacy challenges."

Where is the legal boundary?

"Companies do not entirely lack the technical ability to strengthen protection, but everyone weighs the real-world costs." Zhong Junhao told reporters bluntly that, for example, strict occlusion detection is prone to misjudgment, resulting in situations where "it cannot shoot when it should," interfering with ordinary users' normal recording.

In addition, in the absence of mandatory standards, if one company takes the lead in adding hardware protection, it may instead be interpreted by the market as having many product restrictions and a poor shooting experience, leaving companies lacking the motivation to take the initiative to add more.

"For users who buy the products, being secretly photographed is a low-probability event, but the increased costs and experience constraints brought by strengthened privacy protection are certain costs they must bear," Zhong Junhao said. The ultimate goal of privacy protection for smart glasses is to achieve: turning "I am always watching" into "you always know I am watching." All-day standby does not equal boundless collection; perception and recording must be strictly separated. Hardware prompts, local storage, legal regulation, and platform governance require multi-party coordination so that "trustworthy" can become the underlying tone of the AI glasses industry.

On product design compliance, Yang Xun suggested that when photographing an individual's portrait, the person being photographed should be able to reasonably identify it. For example, if AI glasses capture relatively sensitive information, there should be more obvious prompts or warnings. If the most sensitive matters are involved, such as military restricted zones or other places where photography is prohibited, there should be an automatic blocking function.

Hu Yanping, a distinguished professor at Shanghai University of Finance and Economics, said bluntly that video recording with AI glasses is itself a highly intrusive function. Even when the indicator light is on, other people in the user's surrounding environment often do not easily notice it and may be photographed without knowing. If the indicator light is covered, shooting becomes more covert, and basically the vast majority of users who use opaque stickers to cover the indicator light are almost without exception doing so to take private photos without others' permission, or their direct purpose is secretly photographing.

Hu Yanping believes that AI glasses have become a disaster area for online secret photography, privacy issues, and portrait infringement. In response, AI glasses manufacturers should not add fuel to the fire or tacitly acquiesce, but should optimize product design, clarify usage rules, follow mandatory standards, make indicator lights and other shooting prompt functions more obvious, and ensure that when the indicator light is covered, the photo and video shooting functions cannot be activated. In addition, policies and legislation are needed to restrict the places and functional uses of AI glasses, for example, secret photography without others' permission should not be allowed in public places, and ordinary people's privacy and dignity, like portrait rights, need to be protected.

"Artificial intelligence is not a place beyond the law. Smart products such as AI glasses can become innovative applications that convenience hundreds of millions of users, but they must not become tools for some people to do whatever they want or even commit crimes," Hu Yanping said. Next, it is necessary to clarify the data boundaries, privacy boundaries, and behavioral boundaries of AI glasses, as well as obligations regarding shooting prompt notification, through industry standards (even mandatory standards), product specifications, usage rules, and policy legislation. Intelligence is boundless, but rights have boundaries.

(Source: The Paper) Middle of the article Embedded advertisement