Meta's AI glasses turned every creep into a body cam, and now the company can't moderate its way out of the mess it built.
The Summary
- Instagram promised to remove harassing videos filmed with Meta's AI glasses after boss Adam Mosseri announced a crackdown. A month later, dozens remain live on the platform.
- Meta removed only nine videos after Business Insider flagged 20-plus examples from verified accounts. The company claims thousands have been taken down overall and search terms like "rizz" are now blocked.
- The gap between Meta's enforcement rhetoric and reality shows what happens when hardware outpaces moderation infrastructure.
The Signal
Meta shipped a product that lets anyone record strangers without visible cameras, then acted surprised when people used it exactly as predicted. The Ray-Ban Meta glasses have a tiny LED indicator that lights up during recording, but in daylight or from certain angles, it's invisible. Perfect for pickup artists who want footage of women's reactions without consent.
Content creator Colin Allen's "secret code" video is textbook: approach a stranger, deliver a scripted line, capture her confusion for engagement metrics. Meta removed it only after being directly alerted. Allen's other videos using the same line on different women stayed up.
"Meta removed the video and eight others after being alerted. The rest remain live."
This is not a moderation problem. This is a product design problem masquerading as a content policy problem. Meta built glasses that make covert recording trivial, marketed them as lifestyle tech, and now runs whack-a-mole with the predictable flood of harassment content. The company says thousands of pieces have been removed, but won't say how many remain or how detection works.
The enforcement is uneven because the rules are unenforceable at scale. What counts as harassment versus "lifestyle and comedy" content? Allen describes himself as the latter, not a pickup artist. Meta has to make judgment calls on tone, context, and intent for every video. That's expensive human labor or unreliable AI moderation. Meanwhile, creators optimize around the edges.
Key moderation gaps:
- Blocking "rizz" in search doesn't remove videos already posted or stop new uploads with different tags
- Verified accounts get more leeway, intentionally or not
- Videos get flagged only after going viral or being reported by press
The real signal: wearable AI cameras are moderation hell. You can't detect what you can't see being recorded. By the time the video hits Instagram, the damage is done. The woman in the tank top didn't know she was being filmed until the video went live. That's the feature, not the bug, for engagement farmers.
The Implication
Every company racing to ship AI-powered wearables needs to solve consent infrastructure before hardware launch, not after. Meta's approach, reactive moderation on uploaded content, guarantees harm happens first and enforcement happens maybe. Expect this pattern to repeat as OpenAI, Google, and startups ship their own recording devices.
For users: if someone approaches you wearing Ray-Ban Meta glasses and starts a conversation that feels scripted, you're probably being recorded. The LED indicator is not reliable protection. The real test of these products is whether companies design them for the world they want or the world they know they'll get.