Private Eyes – The Legal Battle over Meta’s Ray Ban Vision
The Multimodal Promise
Smart glasses give the user a number of opportunities and Meta’s (FB) Ray-Bans (with display) are the quintessential product. Depending on the model, you can look at an object and ask the AI, “What is that?”. The Ai can see what you see and figures out what the object is and provides details. Of course you can take it even further and ask where it can be purchased, whereupon the glasses will give you visual ‘cards’ with store names and addresses, and in some cases you can purchase the item directly through Instagram Shopping or Facebook Marketplace, although the actual checkout would be on your phone.
Capture Mechanics and the Transparency Light
Users can also take images or video while wearing the glasses, and to make it simple for the user, there are three ways in which to do this.
Super Sensing: From Active Inquiry to Passive Observation
However, Meta has recently introduced “Super Sensing” (aka “Live AI”), which enables a more continuous flow of data if you have it enabled. In fact, the glasses take a picture every few seconds so the user can have an ongoing conversation with the AI about what is being seen, without having to re-activate the AI “Hey Meta”. The recording light stays on or pulses during this mode, but it seems that sometimes users forget or don’t notice that the glasses are still taking periodic snapshots, even after they put the glasses down.
The Human-in-the-Loop: Data Annotation in Kenya
As the snapshots in this mode are being sent to the cloud, Meta has been using some of these images and videos to train its AIs, and that means that some of the images have to be reviewed by human contractors in order that they can be labeled for the AI (Yellow dog on green grass, green boat in blue ocean, etc.). A recent controversy has developed over the fact that these 3rd party contractors are therefore viewing what is personal data, which has put privacy advocates up in arms. A major investigation by Swedish newspapers (Svenska Dagbladet and Göteborgs-Posten) just last month revealed that human contractors in Kenya have been reviewing thousands of "accidental" recordings. These include:
This can happen in two ways:
While there is a difference between legal ownership and usage rights, the reality is that to the user, they are the same. You, as the creator of the image or video, own a US copyright and Meta agrees that you retain ownership of your “User Content”. However when you put on the glasses and use the cloud-based AI features, you give Meta a “License” to everything you create, along with its partner Luxottica (pvt). "perpetual, non-exclusive, royalty-free, transferable, and sub-licensable worldwide license", which means they can store it on their servers, process it for their AI models, share it with 3rd party contractors, and use it to improve their AI models.
Meta updated its policy in January to explicitly state that interactions with Meta AI (including the visual data from your glasses) can now be used to personalize ads across Facebook and Instagram. This “intent” data can be used to personalize ads in the future and also allows the data to be reviewed by humans and even if you delete your activity logs, the insights generated in your personal advertising profile will remain. Meta does let you opt out of some things. You can turn of the storage of AI interactions, and you could always disable cloud processing, but that would remove much of what the glasses are able to provide, which leads most to scroll through the privacy notice and say yes to it all.
The Legal Challenge: Challenging "Designed for Privacy"
That said, not everyone feels that way, and last week Meta received a proposed class action lawsuit in the US District Court of the Northern District of California – San Francisco. The lawsuit seeks to hold Meta responsible for what it alleges is “affirmatively false advertising and failure to disclose the true nature of surveillance and its connection to the company’s AI data collection pipeline”. It further alleges that Meta’s conduct violates state consumer protection laws and offends basic notions of privacy.
The underlying issue in the suit is Meta’s advertising motto for the glasses, “Designed for privacy, controlled by you”. The suit points to product privacy pages in ads and product description that state, “You’re in control of your data and content” and “Clear, easy device and app settings” to “help you manage your information, giving you control over what content you choose to share with others, and when.” Further, pointing to the Meta AI Glasses website, which offers tips for consumers on “How to wear your Ray-Ban Meta glasses responsibly,” in order to make others feel “safe and comfortable while you’re wearing your glasses,”
The suit points to how Meta feeds data, captured by consumers through their Meta AI Glasses, to a subcontractor. At the subcontractor’s headquarters in Nairobi, Kenya, thousands of people working as “data annotators” review this sensitive user data, annotating it with labels like “cars,” “lamps,” and “people” in order to train artificial intelligence used in Meta products. These annotators report seeing deeply private video clips, including videos of bathroom visits, sex, and other personal moments.
One person recounted seeing a video wherein a man placed his Meta AI Glasses on a bedside table and left the room. Afterwards, his wife entered the room and, unknowingly, changed clothes in front of the Meta AI Glasses. The video was sent to people halfway around the world to view and annotate in order to train AI, all without the woman’s knowledge that the video of her had been captured in the first place. Other data annotators report seeing videos with visible bank cards, private text exchanges, and other private information. This sensitive data lives in Meta’s database, viewed by human data annotators and is available indefinitely for use to Meta’s AI products.
The plaintiffs are asking for an injunction, a disgorgement of profits, and punitive damages, as one would expect, but this is less an AI oriented issue and more an issue of how strong Meta’s disclaimers and privacy policies are. We expect the real challenge will come from the ‘implied’ limitations in the literature that accompanies the glasses. If that verbiage is too broad the plaintiffs would have at least a case for misrepresentation, but more likely Meta will add additional wording to the product disclaimers, and perhaps add more “Please acknowledge” boxes that covers the 3rd party viewing of ‘accidental’ personal information.
Conclusion: The "Privacy-by-Design" Paradox
The emerging legal battle in the Northern District of California highlights a fundamental friction between AI utility and user expectation. At its core, the lawsuit suggests that Meta’s "Privacy-by-Design" marketing may have created a false sense of digital isolation. While the hardware includes visible indicators like the white LED, the software architecture relies on a "human-in-the-loop" pipeline that many users find incompatible with the intimate settings in which the glasses are worn.
The outcome of this litigation likely hinges on three critical factors:
Smart glasses give the user a number of opportunities and Meta’s (FB) Ray-Bans (with display) are the quintessential product. Depending on the model, you can look at an object and ask the AI, “What is that?”. The Ai can see what you see and figures out what the object is and provides details. Of course you can take it even further and ask where it can be purchased, whereupon the glasses will give you visual ‘cards’ with store names and addresses, and in some cases you can purchase the item directly through Instagram Shopping or Facebook Marketplace, although the actual checkout would be on your phone.
Capture Mechanics and the Transparency Light
Users can also take images or video while wearing the glasses, and to make it simple for the user, there are three ways in which to do this.
- Press the “Capture” button on the right are of the glasses and hold for a second or two
- You can say “Hey Meta, record a video”
- If you are using the Meta neural band you can double tap your index finger to your thumb
Super Sensing: From Active Inquiry to Passive Observation
However, Meta has recently introduced “Super Sensing” (aka “Live AI”), which enables a more continuous flow of data if you have it enabled. In fact, the glasses take a picture every few seconds so the user can have an ongoing conversation with the AI about what is being seen, without having to re-activate the AI “Hey Meta”. The recording light stays on or pulses during this mode, but it seems that sometimes users forget or don’t notice that the glasses are still taking periodic snapshots, even after they put the glasses down.
The Human-in-the-Loop: Data Annotation in Kenya
As the snapshots in this mode are being sent to the cloud, Meta has been using some of these images and videos to train its AIs, and that means that some of the images have to be reviewed by human contractors in order that they can be labeled for the AI (Yellow dog on green grass, green boat in blue ocean, etc.). A recent controversy has developed over the fact that these 3rd party contractors are therefore viewing what is personal data, which has put privacy advocates up in arms. A major investigation by Swedish newspapers (Svenska Dagbladet and Göteborgs-Posten) just last month revealed that human contractors in Kenya have been reviewing thousands of "accidental" recordings. These include:
- Intimate Moments: Footage of people in bathrooms, getting undressed, or in bedrooms.
- Sensitive Data: Clear images of bank cards, ATM PIN entries, and computer screens with private emails.
This can happen in two ways:
- Session Persistence – Sometimes the “Super” session does not end correctly and continues to record snapshots periodically.
- External Triggers – As the glasses are always listening, background noise can accidentally trigger a session.
While there is a difference between legal ownership and usage rights, the reality is that to the user, they are the same. You, as the creator of the image or video, own a US copyright and Meta agrees that you retain ownership of your “User Content”. However when you put on the glasses and use the cloud-based AI features, you give Meta a “License” to everything you create, along with its partner Luxottica (pvt). "perpetual, non-exclusive, royalty-free, transferable, and sub-licensable worldwide license", which means they can store it on their servers, process it for their AI models, share it with 3rd party contractors, and use it to improve their AI models.
Meta updated its policy in January to explicitly state that interactions with Meta AI (including the visual data from your glasses) can now be used to personalize ads across Facebook and Instagram. This “intent” data can be used to personalize ads in the future and also allows the data to be reviewed by humans and even if you delete your activity logs, the insights generated in your personal advertising profile will remain. Meta does let you opt out of some things. You can turn of the storage of AI interactions, and you could always disable cloud processing, but that would remove much of what the glasses are able to provide, which leads most to scroll through the privacy notice and say yes to it all.
The Legal Challenge: Challenging "Designed for Privacy"
That said, not everyone feels that way, and last week Meta received a proposed class action lawsuit in the US District Court of the Northern District of California – San Francisco. The lawsuit seeks to hold Meta responsible for what it alleges is “affirmatively false advertising and failure to disclose the true nature of surveillance and its connection to the company’s AI data collection pipeline”. It further alleges that Meta’s conduct violates state consumer protection laws and offends basic notions of privacy.
The underlying issue in the suit is Meta’s advertising motto for the glasses, “Designed for privacy, controlled by you”. The suit points to product privacy pages in ads and product description that state, “You’re in control of your data and content” and “Clear, easy device and app settings” to “help you manage your information, giving you control over what content you choose to share with others, and when.” Further, pointing to the Meta AI Glasses website, which offers tips for consumers on “How to wear your Ray-Ban Meta glasses responsibly,” in order to make others feel “safe and comfortable while you’re wearing your glasses,”
The suit points to how Meta feeds data, captured by consumers through their Meta AI Glasses, to a subcontractor. At the subcontractor’s headquarters in Nairobi, Kenya, thousands of people working as “data annotators” review this sensitive user data, annotating it with labels like “cars,” “lamps,” and “people” in order to train artificial intelligence used in Meta products. These annotators report seeing deeply private video clips, including videos of bathroom visits, sex, and other personal moments.
One person recounted seeing a video wherein a man placed his Meta AI Glasses on a bedside table and left the room. Afterwards, his wife entered the room and, unknowingly, changed clothes in front of the Meta AI Glasses. The video was sent to people halfway around the world to view and annotate in order to train AI, all without the woman’s knowledge that the video of her had been captured in the first place. Other data annotators report seeing videos with visible bank cards, private text exchanges, and other private information. This sensitive data lives in Meta’s database, viewed by human data annotators and is available indefinitely for use to Meta’s AI products.
The plaintiffs are asking for an injunction, a disgorgement of profits, and punitive damages, as one would expect, but this is less an AI oriented issue and more an issue of how strong Meta’s disclaimers and privacy policies are. We expect the real challenge will come from the ‘implied’ limitations in the literature that accompanies the glasses. If that verbiage is too broad the plaintiffs would have at least a case for misrepresentation, but more likely Meta will add additional wording to the product disclaimers, and perhaps add more “Please acknowledge” boxes that covers the 3rd party viewing of ‘accidental’ personal information.
Conclusion: The "Privacy-by-Design" Paradox
The emerging legal battle in the Northern District of California highlights a fundamental friction between AI utility and user expectation. At its core, the lawsuit suggests that Meta’s "Privacy-by-Design" marketing may have created a false sense of digital isolation. While the hardware includes visible indicators like the white LED, the software architecture relies on a "human-in-the-loop" pipeline that many users find incompatible with the intimate settings in which the glasses are worn.
The outcome of this litigation likely hinges on three critical factors:
- The "Informed Consent" Standard: Meta will argue that its terms of service technically disclose the possibility of human review. However, the plaintiffs are betting that "buried" disclosures cannot override a high-visibility marketing campaign that explicitly promised users were "in control."
- The Failure of Anonymization: Evidence that facial blurring and data filtering frequently fail, especially in the low-light or high-motion environments where "accidental" recordings occur. This can undermine Meta's primary defense that they take reasonable steps to protect identity.
- The "Super Sensing" Liability: The continuous nature of "Live AI" creates a unique risk where the device transitions from a purposeful tool to a passive observer. If the court finds that "session stickiness" or "external triggers" led to the capture of intimate data without a clear affirmative act from the user, Meta may face significant punitive damages.
RSS Feed