Two senators asked Meta who watches the footage from its glasses. Meta has until Thanksgiving to find out
Meta sells camera glasses with an LED that is supposed to tell bystanders they are being recorded. reports and a lawsuit say contractors reviewing AI footage from those glasses saw people undressing, using the bathroom and having sex. on October 28, Senators Ron Wyden and Ed Markey sent Mark Zuckerberg a letter with questions that Meta could have answered before it shipped. the sloth has a few to add.
what the letter actually asks
stripped of the Senate letterhead, the questions boil down to:
- the LED. how often does the camera activate without the wearer meaning it to, and what does Meta do about people who tamper with or cover the recording indicator?
- bystanders. does Meta blur faces and license plates in imagery processed for certain AI interactions, the way it says it does on Facebook and Instagram?
- human review. under what circumstances is imagery from the glasses made accessible to employees or contractors, and when is it used to train AI models?
- deletion. what rights do users and the people around them have to get that imagery deleted?
- due diligence. citing Meta’s own documents, did Meta conduct detailed human rights due diligence and safety testing before release, given it had already identified bystander privacy risks?
the senators’ line, per Engadget: “Meta’s professed commitment to privacy and preventing intimate image abuse rings hollow if the company cannot demonstrate meaningful steps to prevent its own products and practices from facilitating those harms.”
the hypocrisy section, as promised by the category name
the letter is the good part. the context around it is where things get funny, in the bleak sense.
- the bystander never agreed to anything. Meta’s disclosures, per 404 Media’s reporting, allow imagery processed for visual AI features to be retained for training, sometimes with human review. the wearer clicked a terms of service. the person in the bathroom stall did not.
- California had a shot. earlier in October, Governor Gavin Newsom vetoed a bill that would have banned using wearables to secretly record people in private areas of businesses, per Engadget. the state that loves to regulate tech found this one a bridge too far.
- the government is also wearing them. the glasses have been spotted on ICE agents during raids, and the NYPD’s counterterrorism unit reportedly raised surveillance concerns about the devices in January. it is hard to write privacy rules for a gadget your own agents find handy.
- firmware decides consent now. users have reported the glasses recording conversations without consent after a firmware update. consent that can be changed by an over-the-air push is a setting, not consent.
and the classic: the same industry that tells Congress self-regulation works is shipping a camera on your face whose main safety control is a small light anyone can cover with a sticker.
the security people’s version of this story
strip out the politics and this is a data flow problem every CISO knows:
- collection you cannot see. a sensor captures people who are not your users and cannot opt out.
- purpose creep. imagery collected to answer “what am I looking at” ends up in a training pipeline.
- humans in the loop. labeling and quality review means contractors, often third party, often offshore, looking at raw data. that is a classic insider and vendor risk surface, and the most sensitive data in the pipeline is exactly what reviewers are most likely to see, because it confuses the model.
- retention without deletion. once frames are in a training set, “delete my data” becomes a research problem, not a database query.
if a vendor pitched that architecture for patient data, it would not pass a first-round review. it shipped as a consumer product, because consumer privacy in the US is mostly a vibe enforced by angry letters.
what an angry letter can and cannot do
a letter from two senators has no legal force. it creates a paper trail, gives reporters a deadline and puts on record what Meta says it does. if the answers later turn out to be wrong, that matters for regulators like the FTC, which has an existing consent order history with Meta over privacy. what it does not do is change a single retention setting tomorrow. federal privacy law is still not a thing, and state wearable recording bills are getting vetoed.
if you run a business, here is the practical bit: you can decide what happens on your premises. a lot of medical offices, law firms, gyms and HR departments already ban recording devices in sensitive areas. smart glasses are recording devices that look like glasses. update the policy and tell staff it covers them.
the sloth take
Meta built a camera that sends what it sees to an AI pipeline with human reviewers, then acted surprised that the footage included the most private moments of people who never agreed to be filmed. two senators want answers by Thanksgiving. the sloth predicts a response letter full of the words “industry-leading”, “privacy by design” and “we take this seriously”, and zero changes to what reviewers can see. prove the sloth wrong, Meta. it would be a first.
more of the same energy on the site: politics and policy pieces live under Politics & Hypocrisy, and the scam side of consumer tech is in task gig job crypto scams and the FTC.
sources
- 404 Media: Senators Demand Meta Answer for ‘Pervert Glasses’ Following 404 Media Investigation
- Engadget: Senators are looking into ‘serious privacy risks’ with Meta’s AI glasses (Oct 30, 2026)
- VR.org: Wyden and Markey gave Meta until November 25 to answer nine questions
- Bloomberg Government: Meta pressed by Democrats to address concerns over AI glasses
