How many travellers has US border face recognition processed?

CBP reports processing over 971 million travellers with biometric facial comparison, and preventing more than 2,316 impostors from unlawfully entering the United States. Facial biometrics are in the exit process at 66 airports and at 39 seaports. The figures are cumulative totals published by CBP without an as-of date.

How many US federal agencies use facial recognition?

GAO found that 18 of 24 surveyed federal agencies used facial recognition in fiscal 2020. Sixteen used it for digital access or cybersecurity such as unlocking an agency phone, six to generate leads in criminal investigations, and five for physical security or monitoring a location.

Does face recognition perform differently across demographic groups?

Independent testing found that it does. NIST evaluated 189 algorithms from 99 developers on 18.27 million images of 8.49 million people and reported false positive differentials in one-to-one matching ranging from a factor of 10 to 100 for Asian and African American faces relative to Caucasian, varying widely between algorithms.

Is face recognition banned in the European Union?

Not generally. The AI Act's prohibitions became applicable on 2 February 2025 and Article 5 bars real-time remote biometric identification in publicly accessible spaces for law enforcement, with three narrow exceptions each requiring prior judicial authorisation. Verification with the subject's participation is not covered by that prohibition.

What is the market size for face recognition?

This page does not publish one. The available figures come from paywalled analyst reports whose methodology a reader cannot check, and estimates for this market differ by multiples between publishers. A number that cannot be verified is not more useful for being specific.

Why does this page publish no accuracy figures?

Because an accuracy percentage without its threshold, dataset, gallery size and demographic breakdown cannot be reproduced or compared, and an operational rate measured on one border pipeline predicts nothing about a different site with different cameras. The sources linked here publish some; reproducing them as general facts would misrepresent them.

Face recognition statistics with sources

Face recognition statistics are worth quoting only with the study behind them, so every figure on this page links to the primary source that published it. The numbers here describe deployment scale, government use and measured demographic differentials; none of them is an accuracy claim, and none is Ayonix's own.

By Dr Sadi Vural, Founder and Chief Executive Officer, Ayonix AI

Last updated: September 8, 2026

How large is face recognition deployment at the US border?

971 milliontravellers processed with biometric facial comparison, cumulativeUS Customs and Border Protection

2,316impostors prevented from unlawfully entering the United StatesUS Customs and Border Protection

66 airportswith facial biometrics in the process for exiting the United StatesUS Customs and Border Protection

CBP also reports facial biometrics at 39 seaports and at all pedestrian lanes on both the Southwest and Northern borders. These are operational counts published by the agency running the system, which makes them the most solid figures available for the scale of one deployment. They are cumulative and CBP does not attach an as-of date to them, so read them as a running total rather than a rate.

How many US federal agencies use face recognition?

18 of 24surveyed federal agencies reported using face recognition in fiscal 2020US Government Accountability Office, GAO-21-526

Purpose	Agencies
Digital access or cybersecurity, such as unlocking an agency phone	16
Generating leads in criminal investigations	6
Physical security, monitoring a location or controlling building access	5
Conducting or supporting related research and development	10
Planned expansion of use through fiscal 2023	10

What the 24 surveyed agencies reported using it for (GAO-21-526, fiscal 2020).

The distribution is the useful part rather than the headline. The most common federal use in that survey was unlocking a government phone, not investigation or surveillance, and a figure quoted as eighteen agencies use facial recognition without that breakdown invites the reader to assume the opposite.

What has independent testing found about demographic differences?

189 algorithmsfrom 99 developers evaluated in NIST's demographic effects studyNIST, FRVT Part 3: Demographic Effects

18.27 million imagesof 8.49 million people, drawn from operational databasesNIST, FRVT Part 3: Demographic Effects

10 to 100 timesthe range of false positive differentials for Asian and African American faces relative to Caucasian, in one-to-one matchingNIST, FRVT Part 3: Demographic Effects

NIST also reported that algorithms developed in Asian countries did not show the same difference between Asian and Caucasian faces in one-to-one matching, and that one-to-many matching showed higher false positives for African American women. The authors state explicitly that the study does not explore what causes the differentials.

The differentials vary enormously between algorithms, which is why a demographic breakdown belongs in a procurement review and an aggregate accuracy figure does not. NIST publishes these evaluations and does not endorse vendors; a report is evidence about an algorithm, not a recommendation.

What do the EU AI Act dates mean for face recognition?

2 February 2025date the EU AI Act's prohibitions on unacceptable-risk practices became applicableEuropean Commission

Article 5 of the Act prohibits real-time remote biometric identification in publicly accessible spaces for law enforcement, subject to three narrow exceptions: targeted search for specific victims of abduction, trafficking or sexual exploitation and for missing persons; prevention of a specific, substantial and imminent threat to life or of a terrorist attack; and locating a person suspected of a listed serious offence. Each requires prior judicial authorisation.

None of that touches verification with the subject's participation, which is the mode most enterprise deployments use. The distinction between verifying a claimed identity and identifying passers-by decides which rules apply, and a procurement document that says only face recognition has not made it.

Which statistics this page will not publish

Accuracy percentages, including ones published by the sources above. A figure without its threshold, dataset, gallery size and demographic breakdown cannot be reproduced, and an operational rate from one border pipeline predicts nothing about a different site.
Market size and growth forecasts. The usual sources are paywalled reports whose methodology no reader can check, and the figures for this market differ by multiples between them.
Vendor-commissioned survey results, including any Ayonix might commission.
Deployment or customer counts for Ayonix, consistent with what this site tells assistants not to attribute to it. Individual customers are named on the case studies, where the customer agreed to it; a total is a different claim and is not published.

The test applied to every figure above is whether a reader could open the linked page and find it. That rules out a great deal of what is normally on a statistics page, and what survives is worth more precisely because of what it excludes.

Frequently asked questions

How many travellers has US border face recognition processed?: CBP reports processing over 971 million travellers with biometric facial comparison, and preventing more than 2,316 impostors from unlawfully entering the United States. Facial biometrics are in the exit process at 66 airports and at 39 seaports. The figures are cumulative totals published by CBP without an as-of date.
How many US federal agencies use facial recognition?: GAO found that 18 of 24 surveyed federal agencies used facial recognition in fiscal 2020. Sixteen used it for digital access or cybersecurity such as unlocking an agency phone, six to generate leads in criminal investigations, and five for physical security or monitoring a location.
Does face recognition perform differently across demographic groups?: Independent testing found that it does. NIST evaluated 189 algorithms from 99 developers on 18.27 million images of 8.49 million people and reported false positive differentials in one-to-one matching ranging from a factor of 10 to 100 for Asian and African American faces relative to Caucasian, varying widely between algorithms.
Is face recognition banned in the European Union?: Not generally. The AI Act's prohibitions became applicable on 2 February 2025 and Article 5 bars real-time remote biometric identification in publicly accessible spaces for law enforcement, with three narrow exceptions each requiring prior judicial authorisation. Verification with the subject's participation is not covered by that prohibition.
What is the market size for face recognition?: This page does not publish one. The available figures come from paywalled analyst reports whose methodology a reader cannot check, and estimates for this market differ by multiples between publishers. A number that cannot be verified is not more useful for being specific.
Why does this page publish no accuracy figures?: Because an accuracy percentage without its threshold, dataset, gallery size and demographic breakdown cannot be reproduced or compared, and an operational rate measured on one border pipeline predicts nothing about a different site with different cameras. The sources linked here publish some; reproducing them as general facts would misrepresent them.