Meta just permanently disabled cameras on thousands of its AI glasses for light tampering — then got sued for allegedly using those same devices to vacuum up training data without consent.
The Summary
- Meta bricked cameras on thousands of AI glasses after detecting users who drilled out or covered the recording indicator lights — less than 0.1% of devices, but potentially 7,000 pairs based on 2025 sales figures
- New privacy lawsuit alleges Meta used its glasses to train AI models on footage captured without proper consent, dubbing them "perv glasses"
- The crackdown and lawsuit arrive weeks after Meta agreed to the largest child-safety settlement in tech history
- Signal: Meta faces a credibility paradox — aggressively policing bad actors while standing accused of being one itself
The Signal
Meta sold 7 million AI glasses in 2025 alone. That's millions of always-ready cameras walking around restaurants, gyms, schools, and bedrooms. The company just confirmed it permanently disabled cameras on thousands of those devices after detecting physical tampering with the tiny recording indicator lights. Users drilled them out or covered them up, making covert recording trivial.
The enforcement hit less than one tenth of a percent of all glasses sold. Do the math: if 0.1% of 7 million is 7,000 pairs, that's 7,000 people who cared enough about secret recording to physically modify hardware. The real number of people who wanted to record without consent but didn't modify their glasses? Unknown.
"Meta dropped the hammer on thousands of its AI glasses users who it detected physically tampering with the recording light."
Now comes the lawsuit. Plaintiffs allege Meta used footage from these same glasses to train its AI models without informed consent. The complaint arrived weeks after Meta settled the largest child-safety case in tech history — a settlement amount that remains undisclosed but reportedly set records. The timing matters. Meta is simultaneously playing sheriff and standing accused of digital trespassing at scale.
The "perv glasses" nickname stuck for a reason. Business Insider previously reported on Instagram's struggles to remove harassing content captured with the devices. The glasses make surveillance frictionless. No phone to hold up. No obvious lens. Just a normal-looking pair of Ray-Bans that happen to see everything you see and remember it forever.
Here's what makes this lawsuit different from typical privacy complaints:
- It targets the training data pipeline, not just the initial capture
- It questions whether "feeding the beast" — Meta's AI training apparatus — was the point all along
- It arrives after Meta publicly committed to cracking down on misuse, creating an awkward optics problem
Meta's crackdown shows the company knows exactly how its glasses are being weaponized. The software update that detected tampering proves Meta has always had the technical capability to monitor how people modify and use the devices. Which raises the obvious question: if Meta can detect a drilled-out LED, what else can it detect, log, and analyze?
The Implication
If you're building AI agents that rely on real-world visual data, watch this case. The lawsuit targets the training layer, not just the capture layer. That's a new front in privacy law. If courts decide ambient visual data captured by wearables requires explicit opt-in consent before feeding training models, the entire economics of embodied AI shift overnight.
For Meta, this is a branding crisis dressed up as a legal problem. You can't credibly position yourself as the safety police while facing allegations that you turned millions of users into unwitting data mules. The company needs to decide: are these cameras for users, or are users cameras for the company?