Eyeglasses, AI, ethics: the optical industry at the heart of the debate
In 2024, facial recognition was still a topic of debate in think tanks and law schools.
By September 2026, it's hanging on the noses of millions of people in the form of a pair of Ray-Ban sunglasses. Meanwhile, researchers resign from Anthropic, publicly declaring that humanity could become extinct because of AI. It would seem that the ethical debate has officially changed its dimension…
This is no longer a question of a hypothetical future: the battle lines are drawn today, in our daily lives—and in the optics industry as well.
What facial recognition actually does, technically
Facial recognition relies on a three-step process. A camera—standard, infrared, or 3D—captures a face. An algorithm extracts biometric characteristics: distance between the eyes, nose shape, jawline. This data is converted into a unique digital model (a facial fingerprint) and compared to a database. The accuracy of this process depends directly on the optical quality of the sensor: resolution, lighting management, and capture angle. This is why optics are at the heart of the system.
What has radically changed in recent years is miniaturization. The cameras that once enabled this type of identification required fixed and visible equipment: airport terminals, surveillance cameras. Today, a 12-megapixel camera filming in stabilized 3K fits in the temple of a 36-gram frame. And it's sold at your optician's.
Ray-Ban Meta: When glasses become tools for mass surveillance
This is the most documented case of the year in the optical sector. More than 7 million pairs of Ray-Ban Meta sunglasses were sold in 2025 , three times more than the previous year. They are worn by joggers, students, and tourists, and are available from €247 in partner optical stores.
What is a cool yet practical tech product can, if misused, become a scandal-mongering machine. Among those that have recently emerged in rapid succession:
The I-XRAY project (Harvard, 2024). Two students paired Ray-Ban Meta sunglasses with PimEyes, a commercial facial recognition engine. The result: they identified strangers on the Boston subway, retrieving their names, addresses, phone numbers—and in some cases, social security numbers—in seconds. They approached a woman on the street, pretending to know her. She believed them. The demonstration relied on publicly available tools.
The " NameTag " module (Wired/Clubic, June 2026). Wired revealed the existence of a dormant module in the Ray-Ban Meta app, internally called NameTag : the glasses' camera captures a face, generates a biometric fingerprint, compares it to a database stored on the phone, and alerts the user if there is a match. This code had already been distributed to millions of devices as early as January 2026—without the functionality being visible in the interface. Meta quietly removed this code in June 2026, after Wired's revelation. For the European Union, activating such a function on a person filmed without their knowledge would constitute a direct violation of the GDPR and the AI Act.
Case of the harassment videos (2025-2026). Dozens of videos of women filmed without their knowledge by men wearing Ray-Ban Meta sunglasses circulated on Instagram. Adam Mosseri, Instagram's CEO, promised sanctions at the end of July. A month and a half later, this content was still circulating freely. The LED meant to indicate recording goes unnoticed in broad daylight—and tutorials are circulating on how to cover it, with some technicians even offering to remove it for a fee…
The class action lawsuit (September 4, 2026). Two fathers sued Meta in Illinois federal court on behalf of their minor children, accusing the company of collecting millions of images to train its AI models while secretly developing NameTag. Norway is calling for their ban in public spaces.
The regulatory framework: Europe is ahead, but practices are lagging behind.
The European AI Act , the first provisions of which came into force on February 2, 2025, explicitly prohibits real-time remote biometric identification in public spaces, except for strictly defined security exceptions. This is precisely what NameTag would have allowed. The CNIL (French Data Protection Authority) has therefore taken a stance on discreet data collection via wearable devices! In France, Article 226-1 of the Penal Code punishes the recording of images in a private place without consent with one year of imprisonment and a €45,000 fine.
But the street remains a legal blind spot. Filming a person alone without their consent is illegal under image rights and the GDPR. However, proof is difficult to establish, and the recording LED, according to the CNIL itself, has a "range that is sometimes limited over long distances." Technology is moving faster than the law…
The warning from within: when the creators of AI sound the alarm
This is where the ethical debate shifts into another dimension. On September 9, 2026, Jacob Coxon , 27, a former pre-training researcher at OpenAI and then Anthropic, resigns and publishes a message that will be shared millions of times.
Speaking on the BBC, he stated: "If we do not slow down the current pace, there is a strong probability that we could all die in the near future."
He adds that some of his colleagues are considering "buying land somewhere because they are so afraid of the instability due to the rapid progress of AI."
This is not an isolated voice. Evan Hubinger , an executive at Anthropic in charge of alignment, publicly estimates the probability that AI will kill all humans in the next decade at over 10% .
Samuel Marks , another security researcher, stated that "the more senior the employee, the more concerned they are".
And in February 2026, Mrinak Sharma , former head of research on protection mechanisms at Anthropic, had already left the company, leaving a similar warning letter.
On September 12, 2026, the day before Coxon's media appearance, Dario Amodei , CEO of Anthropic , published an essay entitled "We Must Pace the Frontier," calling on the industry to slow down and announcing that Anthropic would grant third-party evaluators permanent access to its systems.
These warnings are a subject of debate.
Clément Delangue ( Hugging Face ) called them "completely absurd." Other observers point to a possible communication strategy: Anthropic is preparing for an IPO potentially valued at $2 trillion, and calls for regulation may also serve to curb competitors.
But in 2023, hundreds of experts had already signed a " declaration on the risk of extinction linked to AI ," calling for this issue to be treated as seriously as pandemics or nuclear weapons. ( Le Devoir, September 2026 ) Coxon's stance is part of a long line of warnings.
AI, facial recognition, law and alerts: what about the optical sector?
The convergence between smart glasses, facial recognition and generative AI puts the optical industry in an unprecedented position.
Products sold in stores, at approved opticians, with possible visual correction, now include cameras, microphones, cloud connections, and potentially biometric identification capabilities.
This is not an abstract question.
It's a matter of professional responsibility, advising the client, and soon, regulatory compliance. Informing a client of the legal implications of a pair of Ray-Ban Meta v sunglasses is now part of the optician's role, just like explaining the difference between photochromic and polarized lenses.
Our question posed in 2024: "Are we ready to sacrifice anonymity for greater security and comfort? " found a partial answer in 2026: many people are ready, often without fully understanding the consequences. The optical industry, however, must be aware of this and support it.
Featured Photo Credits: Samon Yu and Will Chen
