Privacy and Opt-in Face Recognition with AI Avatars

Why this question comes up

Any machine that can "see" a visitor raises a fair question: what happens to that data afterwards? It's one of the first things we get asked, whether the avatar is going into a government service centre, a mall, or a corporate reception, and it deserves a direct answer rather than a vague reassurance buried in a privacy policy nobody actually reads.

Opt-in, not default

Face recognition on a faceOS avatar is opt-in only, in every single placement. The avatar's ability to see when someone is ready to talk — one of her five core abilities, alongside listening, speaking, showing and completing a task — doesn't require identifying who that person is. Recognition features that do identify a visitor are switched on only with that visitor's explicit agreement, not running by default in the background without anyone's knowledge.

The right to be forgotten

Where face recognition is in use, a visitor can ask to be forgotten at any time, and that request is honoured in full. This isn't a one-time consent that locks someone in going forward — it's a standing right a visitor can exercise whenever they choose, for as long as they interact with the avatar, not something they had to opt out of before their first visit or lose the chance to withdraw later on.

What this means in practice on-site

For a business putting an avatar on a 7-foot holobox, 21-inch holobox or 2D tabletop screen, this translates directly into a straightforward setup conversation: you decide whether your use case actually needs recognition at all, and if it does, visitors are asked rather than assumed, and can opt out or ask to be forgotten at any point they choose. For many placements — a reception avatar that simply answers questions and points visitors in the right direction — recognition isn't required at all, since "seeing" that someone wants to talk is a different, narrower thing than identifying who that specific person is.

Where this question comes up most

It's a particularly common question in hospitality and real estate sales centres, where visitors are naturally more aware of being observed in an unfamiliar space, and in government or corporate settings where data handling is already a point of scrutiny for other, entirely unrelated reasons. We go into the practical side of setting up an avatar for guest-facing roles in digital receptionists for hotels and hospitality.

What "seeing" means versus what "recognising" means

It's worth carefully separating two things that often get conflated in this conversation. Seeing that someone is nearby and ready to talk is a basic input every avatar needs to start a conversation naturally, the same way a person glances up when someone approaches their desk — it tells the avatar when to begin, nothing more. Recognising who that specific person is, and linking that to an identity or a history, is a separate and much more sensitive capability — and that's the part that's opt-in, not the first.

Why we lead with a direct answer

Visitors, and the businesses that choose to host an avatar, are entitled to a plain answer on this rather than fine print buried somewhere nobody ever looks. We'd rather state the opt-in rule and the right to be forgotten clearly up front, in a blog post and on our FAQ, than have a business discover the details only when a visitor asks about it on-site in front of other people. If your use case has requirements beyond what's described here, that's a conversation worth having with us before you commit to a venue setup, not after it's already installed.

Ask us, or ask her directly

If you have a specific scenario in mind — a certain kind of venue, a particular regulatory question, a concern raised by your own visitors — the most direct way to get a clear answer is simply to ask. You can also see the opt-in approach explained in the avatar's own words, in the same conversational style she uses for everything else.

Have a specific question about how this works for your own venue? Book a live demo.