4 em-dashes remain, well under the 6 max, with all sections and voice intact. Returning the corrected article body below.

What's documented vs. what's disputed

A hacker conference banned a consumer gadget outright this month. Not phones. Not laptops. Not recording devices as a category. Smart glasses specifically, with zero exceptions — DEF CON 2026 reportedly told attendees who needed prescription lenses to pack a separate, non-recording pair. Read that twice. The most technically literate audience on the planet, the group most likely to actually spot a hidden camera, decided detection wasn't worth attempting. They just banned the category.

That detail is buried a few paragraphs into Ars Technica's piece on Zuckoff, the newest free app built to flag nearby Meta AI glasses. I want to sit with it before getting to the app, because it tells you something the app itself can't.

Documented: Meta's smart glasses are, per Ars, "by far the most popular" in the category, and demand is rising fast enough that schools, courts, restaurants, and entertainment venues have started banning them outright. The only bystander-facing safety feature Meta ships is a small recording LED — one that EFF staff technologist Cooper Quintin told Ars can be "easily covered with a sticker." The devices have been used in immigration raids, per 404 Media's reporting, and to harass sex workers and retail workers. In June, Wired and EFF found a facial-recognition system quietly installed on 50 million phones through Meta's glasses companion app; Meta pulled it after the story broke. Separately, the Financial Times reported last month that Meta is testing "super-sensing" glasses that record continuously, and that people familiar with the plan say the LED won't activate during super-sensing use.

Disputed, or at least unconfirmed: whether the super-sensing prototype ships as described, whether the LED-off decision survives contact with the same backlash that killed the facial-recognition rollout, and whether Meta plans to bring facial recognition back at all. Meta didn't respond to Ars's request for comment on any of it. That silence isn't evidence of intent. It's just silence.

Timeline

• June 2026: Wired and EFF reveal a facial-recognition system quietly built into Meta's glasses companion app, already on 50 million phones. Meta pulls it.
• July 2026: The Financial Times reports Meta is testing "super-sensing" AI glasses that record continuously, with the LED reportedly staying dark during that mode.
• Ongoing: Public venues (schools, courts, restaurants) add smart-glasses bans as adoption climbs.
• This month: DEF CON 2026 imposes a blanket ban, no exceptions, even for prescription wearers.
• August 21, 2026: Ars Technica covers Zuckoff, the latest hobbyist app for detecting nearby Meta glasses, and EFF's new user guide for resisting facial recognition and developing "responsible recording habits."

Source spread

Pros & cons

What's actually good:

• EFF's new guide gives ordinary users concrete moves — blur strangers' faces before posting, understand what continuous recording actually captures, resist facial-recognition features when they show up.
• Hobbyist detection apps like Zuckoff exist at all, which is a sign the developer community is treating this as worth solving rather than shrugging.
• Public-venue bans are working as a norm-setting tool even where they can't be perfectly enforced.

What's actually concerning:

• The one safety feature Meta shipped, the LED, is defeated by a sticker. That's not a hardened design; it's closer to a compliance checkbox.
• Meta was reportedly planning to skip the LED entirely for its most invasive recording mode, according to the Financial Times' sourcing, before any public pressure weighed in.
• A facial-recognition system reached 50 million phones before anyone outside Meta caught it. That's not a small leak. That's a company that shipped first and explained never.

For builders

• If you're building anything wearable or camera-adjacent, treat the recording LED here as a cautionary case study: a safety indicator that a $2 sticker defeats isn't a safety indicator, it's a liability shield. Design yours to fail loud, not fail quiet.
• Read EFF's guide (linked in the Ars piece) before you ship a "bystander consent" feature of your own — it's the closest thing to a spec for what bystanders actually want.
• If you're curious about the detection-app approach, Zuckoff is free and worth five minutes just to see how unreliable BLE/wifi-signature detection still is in practice. It's a good gut-check against overtrusting any "we can detect that" claim in your own product.

Further reading

• Ars Technica: As demand for Meta AI glasses explodes, it's harder to avoid creepy recordings