The pitch is genuinely seductive. A visitor walks up to your lobby kiosk, looks at a camera, and three seconds later a badge prints with their name on it. No typing. No fumbling with a driver's license. No line building up behind them while someone hunts for the right spelling of their host's last name.
Facial recognition at the front desk works. That's not the problem.
The problem is that the moment your kiosk measures the geometry of someone's face, you stopped collecting contact information and started collecting a biometric identifier — and in a growing list of states, that single change puts you under a body of law with per-violation penalties, mandatory written consent, and in one state, the right for any visitor to sue you personally.
Most organizations discover this after the kiosk is already installed.
The line between "a photo" and "a biometric identifier"
This distinction trips up more security teams than any other, so it's worth being precise.
Taking a visitor's photo and printing it on a badge is, in most jurisdictions, an ordinary business record. You've captured an image. You store it. It's personal data, and general privacy law applies, but you're not in biometric territory.
Running that photo through an algorithm that extracts a mathematical representation of facial structure — the distances and ratios that make a face computationally matchable — is different. Illinois' Biometric Information Privacy Act explicitly names "a scan of face geometry" as a biometric identifier. Texas' Capture or Use of Biometric Identifier Act (CUBI) covers the same ground. That template, that faceprint, is what triggers the obligations.
The practical test: can your system take a new photo and match it against a stored record? If yes, you're processing biometrics, whatever your vendor calls the feature on their pricing page.
This matters because vendors market these systems as "touchless check-in," "express arrival," or "smart recognition." None of those phrases appear in any statute. The underlying technology is what gets regulated.
Illinois BIPA: the one with teeth
BIPA has been on the books since 2008 and remains the strictest biometric law in the United States, for one structural reason: it includes a private right of action. Individuals can sue directly. They do not have to show that they lost money, suffered identity theft, or were harmed in any concrete way. The violation itself is the injury.
Statutory damages run $1,000 per negligent violation and $5,000 per intentional or reckless violation. Courts have held that each separate scan without proper consent can count as its own violation — which is how a visitor management deployment with a few thousand check-ins becomes a class action with a number attached that gets the CFO's attention. Meta, Google, and TikTok have all settled BIPA claims; the case law is mature and plaintiffs' firms are well organized around it.
What BIPA actually requires from a private entity before it collects a faceprint:
- Written notice that biometric data is being collected or stored.
- A statement of specific purpose and retention duration — what you're using it for and how long you'll keep it.
- A written release from the individual. Signed consent, obtained before collection.
- A publicly available written retention and destruction policy, with a defined schedule.
- Reasonable security measures, at least equal to how you protect other confidential information.
- No selling, leasing, trading, or otherwise profiting from the biometric data. This one is absolute.
Read requirement three carefully, because it's where most lobby deployments fail. Implied consent does not satisfy BIPA. A visitor walking up to a kiosk that has a camera on it has not consented. A line buried in your terms of service has not obtained consent. A sign on the wall that says "this area is monitored" has not obtained consent. You need an affirmative, informed, written release — and the check-in flow has to be built to capture and store it.
Texas, Colorado, and the widening map
Illinois is not the whole story anymore, and organizations that operate in multiple states can't treat it as an Illinois-only problem.
Texas (CUBI) carries civil penalties of up to $25,000 per violation. There's no private right of action — enforcement runs exclusively through the Attorney General — but the Texas AG's office has been visibly active on biometric and AI-related privacy matters, and a single non-compliant workflow can be counted as multiple violations across capture and retention.
Colorado amended its Privacy Act via HB 24-1130, with the biometric provisions effective July 1, 2025. The amendment imposes informed-consent requirements, a strict deletion schedule, and a written incident-response protocol specific to biometric data. Enforcement sits with the Attorney General and district attorneys; no private right of action.
Washington has its own biometric statute, making it the third state with dedicated legislation rather than provisions folded into a general privacy law.
Beyond those four, roughly two dozen states now impose some requirement on biometric data through broader consumer privacy frameworks — California, Connecticut, Virginia, Oregon, Minnesota, Maryland, Tennessee, and others. The obligations vary in scope, but the direction is consistent and the list gets longer every legislative session.
For a company with sites in five states, this produces an awkward operational reality: the compliance floor is set by your strictest jurisdiction, not your headquarters. A national rollout configured for whatever is convenient in your home state is a rollout configured for your largest exposure.
Why biometric breaches are categorically worse
There's a reason legislatures singled biometrics out rather than folding them into general data-protection rules.
A password that leaks gets rotated. A credit card that leaks gets reissued. A faceprint that leaks is permanent. Your visitors cannot change their facial geometry after your vendor's storage bucket is left public. The harm doesn't expire, and it follows them to every other system that uses the same modality.
That permanence is what justifies the elevated penalties, and it's also the right frame for the internal conversation. The question isn't "can we defend this if we're audited." It's "are we comfortable being the custodian of something our visitors can never replace."
Six questions to ask before you enable face matching
If facial recognition is on your roadmap — or already live — work through these before your next audit does it for you.
1. Where is the template stored, and who can reach it? On-device, in your tenant, or in the vendor's shared infrastructure? Is it encrypted at rest with keys you control? Which vendor staff have access, and is that access logged? Get the answer in writing, in the contract, not from a sales engineer on a call.
2. Can you produce the consent record for a specific visitor on a specific date? Not "we show a consent screen." Can you run a query and return the signed release, the version of the notice they saw, and the timestamp? If the consent isn't retrievable as a record, you cannot prove it in a dispute, which means functionally you don't have it.
3. Does deletion actually delete? Your retention policy says ninety days. Does the template disappear from primary storage, backups, replicas, logs, and the vendor's analytics pipeline at ninety days — or does it disappear from the UI while persisting everywhere else? Ask the vendor to describe the deletion path component by component.
4. What happens when a visitor says no? There must be a working alternative. A visitor who declines the face scan needs a check-in path that is equally functional and not obviously punitive. If declining means standing aside for fifteen minutes, your consent isn't meaningfully voluntary, and that's exactly how it will be characterized later.
5. Who signed off, and against which state's law? Biometric deployment is not purely a facilities decision. Legal and privacy counsel need to review it against every jurisdiction where you operate, and that review should be documented and dated.
6. Do you actually need it? This is the question that gets skipped. Face matching solves a narrow problem: verifying that a returning visitor is who they claim to be, quickly. If your real pain is slow check-in, pre-registration with a QR code solves it without touching biometrics at all — the visitor gets an emailed pass before arrival, scans it at the kiosk, and is badged in under ten seconds. If your real pain is unverified identity, government ID scanning with immediate destruction of the image covers a large share of the risk with a far smaller regulatory footprint.
A meaningful number of biometric deployments are solving a speed problem with an identity tool, and inheriting a decade of privacy liability to save fifteen seconds.
What a defensible deployment looks like
For organizations that have a genuine need — high-security facilities, sites with repeat contractor traffic, environments where credential sharing is a real threat — biometrics can be run responsibly. The pattern looks like this:
- Consent before capture, as a blocking step. The camera does not activate until the release is signed. Notice text is plain-language, states the purpose and retention period, and is versioned so you know which text each visitor saw.
- Shortest viable retention. If contractors return weekly, you don't need a two-year template. Tie retention to the actual access relationship and delete on a schedule that runs automatically rather than on request.
- A visible, equal alternative. Offer it on the same screen, in the same type size, without friction penalties.
- Segregated, encrypted storage with access logging, and a documented incident-response protocol specific to biometric data — which Colorado now requires outright.
- A published retention and destruction policy, actually reachable on your website, not a PDF in a shared drive.
- Per-jurisdiction configuration. Your Illinois lobby and your Nevada lobby may need different flows. A visitor management platform that can't vary consent requirements by site will force you into the strictest configuration everywhere — which is safe, but often more restrictive than the business wants.
- Contractual clarity on the vendor's role. Who is the controller and who is the processor. What the vendor may do with the data. What happens to the templates when you terminate. Get it in the agreement.
The honest summary
Facial recognition in visitor management is not a bad technology, and it's not going away. It's a technology whose regulatory cost is invisible at purchase and becomes very visible during litigation or an AG inquiry.
The failure mode isn't a company deliberately mishandling faceprints. It's a facilities manager buying a kiosk with a feature enabled by default, no legal review, a consent screen that a lawyer would not call consent, and a retention setting nobody changed from the vendor's default. Eighteen months and forty thousand check-ins later, that's the exposure.
Before you turn the feature on, ask whether you need biometric matching or whether you need faster check-in. They are different problems. One of them has a solution that doesn't require you to become the permanent custodian of thousands of people's faces.
If the answer is that you genuinely need it — build the consent, retention, and deletion machinery first, then enable the camera. Doing it in that order costs a few weeks. Doing it in the other order is how the interesting settlements happen.
KyberAccess provides visitor management with configurable consent capture, per-site policy controls, pre-registration, and ID verification with immediate image destruction — so you can match your check-in flow to your jurisdiction instead of the other way around.
This article is general information about a fast-moving area of law, not legal advice. Consult qualified counsel before deploying biometric technology.