Meta wants its AI glasses to seem less creepy. Its AI strategy says otherwise.
Digital Frontier EditorialJuly 8, 20264 min read
Key Takeaways
Meta admits some users deliberately disable the recording LED on its AI glasses to surveil people — often women — without consent.
The company is simultaneously testing prototypes that continuously capture audio and images every few seconds.
Images shared with Meta AI can train its models, regardless of the "only you" privacy promise.
Kenyan content moderators reportedly viewed graphic sexual and toilet footage from glasses recordings while training Meta's AI.
Meta wants you to believe its AI glasses are safe. Its actions prove otherwise.
The company announced a new safeguard this week: if someone tampers with the LED that signals recording, the camera shuts off. Meta calls this industry leadership. The reality is grimmer. The feature exists because people were already covering the LED with tape to record secretly. When Meta adapted to detect blocked LEDs, determined users developed "sophisticated efforts to modify or destroy the capture LED."
Read that again. Meta is confirming that a subset of its customers uses these glasses as covert surveillance tools. The targets are often women. The company knows this. It built a technical countermeasure instead of asking why its product attracts predators.
The privacy theater
Meta's blog post frames the LED fix as consumer protection. Then it pivots to reassurance: "You, and only you — unless you choose to share them." That promise dissolves on contact with the privacy policy. Any image you share with Meta AI becomes training data. The glasses are a collection funnel disguised as a consumer gadget.
Meanwhile, the Financial Times reports Meta is testing a prototype that continuously collects audio while snapping photos every few seconds. Continuous. Not user-initiated. Not transparent. A wearable panopticon.
The human cost
Kenyan contractors hired to train Meta's AI viewed graphic sexual content and footage of people using toilets — all captured by these glasses. Meta canceled the contract after allegations surfaced. The damage remains. Workers traumatized. Victims recorded without knowledge. Meta's response: sever the vendor relationship, keep the product roadmap.
This pattern repeats. Cambridge Analytica. Biometric settlements. Children's safety failures. Books by whistleblowers document a culture that treats privacy violations as growth expenses. The glasses are not a departure. They are the logical endpoint.
No opt-out from the vision
Meta's AI strategy demands ever more personal data. Training on your images. Enabling AI features by default. Exploring biometric facial recognition. Continuous recording prototypes. Every product decision bends toward collection. The LED fix is a pressure valve — just enough safety theater to keep regulators at bay while the real engine runs unimpeded.
Creepy isn't a reputation problem. It's a business model.