On August 17 a Meta patent application became public, and it travelled fast. Published as US 2026/0238876 A1, “Smart Cameras Enabled by Assistant Systems,” it describes camera glasses that identify the people in the wearer’s line of sight, read expressions and gaze, rank those people by how important they are to the wearer using stored relationship data, capture short clips when something notable happens, and assemble the result into a highlight reel afterwards. The worked example in the filing is a dinner party.
Two things are worth getting right before anyone builds a strategy on this.
It is a published application, not a granted patent. The claims decide the legal scope, and they have not been granted. It is also a continuation of a 2022 filing that continues an application first made in 2019, so this is not a sudden change of direction. Companies patent things they never ship.
But the reason this story has legs is not the legal document. It is that the capability described is no longer speculative, the hardware to run it is already on people’s faces, and nobody in the events industry has decided what to do about it.
Why this lands on events first
Think about where a face-recognizing wearable would be most useful, and most uncomfortable. It is a room full of people you half-know, wearing badges, whose names you are trying to recall, where you will have forty short conversations in two days and remember six of them.
That is a conference. The pitch practically writes itself: never blank on a name again, get a highlight reel of your best hallway conversations, walk away with the moments instead of the business cards.
Which means the first serious deployments of ambient facial recognition are likely to happen inside venues that somebody else is responsible for. Yours.
Event organizers already carry the duty of care for what happens in their rooms. You decide whether sessions are recorded, whether press are admitted, whether a sponsor can film in the expo hall, whether a closed-door roundtable stays closed. Attendee-worn recognition is the same category of decision, arriving through a channel you do not control: the attendee’s own eyewear.
The line that matters is not the technology
It is tempting to read this as a story about facial recognition being creepy. That framing does not survive contact with the events industry, because facial recognition is already all over it and mostly welcome. Attendees upload a selfie and get every photo they appear in. Badge collection uses face verification. Access control uses it at large venues. None of that generates outrage, and it should not.
The distinction that actually predicts whether people accept a system is not the algorithm. It is two questions:
Who initiated it? In a photo-matching hub, the person being recognized is the one who pressed the button. They uploaded their own face, on purpose, to get something they wanted. In the system the patent describes, the person being recognized did nothing at all. Someone else looked at them.
Who gets the benefit? In the first case, the subject gets their own photos. In the second, the subject becomes content in somebody else’s memory reel. That asymmetry, not the underlying model, is what makes a room uneasy.
Every workable policy you can write comes down to those two questions. Hold onto them, because they also apply to the vendors already in your stack.
Three decisions you should make before your next event
1. Take a position on attendee-worn recording, and write it down. There are three defensible stances: permitted, permitted with named exceptions, or prohibited in specific spaces. All three are fine. What is not fine is having no position, because the default then gets set by whoever shows up with the newest hardware, and you find out during the session rather than before it.
2. Decide where recording is off, by space rather than by device. Restrooms, prayer and wellness rooms, medical and nursing rooms, private meeting pods, and any session you sold as confidential. Write the rule about the behavior, not the product. “No recording in this room” still works in three years. “No Ray-Ban Meta glasses” is obsolete the moment somebody ships a competitor.
3. Tell speakers and sponsors separately. They agreed to be in a room. They did not agree to be indexed, clipped, and ranked. A speaker who declines to be recorded by your AV crew has, in practice, declined to be recorded by an attendee’s glasses too, and right now nothing in your run of show reflects that.
We put the full version, with sample registration and signage language you can paste, in a free attendee recording policy checklist.
Now ask your own vendors the same questions
Here is the uncomfortable second half. If ambient recognition raises consent questions, so does every system at your event that touches a face, including the ones you bought.
Most organizers cannot answer basic questions about their own stack. Which vendors create biometric templates rather than just storing images? Is a match initiated by the attendee or by the organizer? Is the comparison scoped to this event, or to everything the vendor holds? How long is biometric data kept, and is that different from how long photos are kept? What happens when someone asks to be erased?
Those are answerable questions with short answers. Ask them in writing, before your next contract renewal, and you will be ahead of most of your peers. The checklist above includes the seven we would ask, and what a real answer sounds like.
Where we stand
We build a facial recognition product, so we are not going to pretend the technology is the problem. Kampfire matches attendees to their own event photos, and it works because the attendee starts it and the attendee is the one who benefits.
What we do think is that the events industry is about to be handed a much harder version of this question by hardware nobody in it chose, and the organizations that already know their own answers will handle it a great deal better than the ones improvising in a session room.
The technology is going to keep getting better at recognizing people. Whether that feels like a service or a violation is decided almost entirely by consent, scope, and who ends up holding the result. Those are your decisions, and they are worth making on purpose.
Start with the attendee recording policy checklist. If you want the plain-language version of how face matching actually works at events, we wrote that too: AI facial recognition technology at events, explained. And if you want to see how we handle attendee data, that is all on our security and compliance page.