United States Customs and Border Protection is moving to give more of its intelligence operation access to Clearview AI, a face recognition tool built around images scraped from the internet. The planned $225,000 agreement covers one year of access and places the technology inside units that collect, compare, and analyze data for security and immigration operations.
The contract matters because it describes a tool intended for regular intelligence work, not only occasional case support. It also arrives while the Department of Homeland Security faces wider scrutiny over how face recognition is being used in federal enforcement activity beyond the border.
What CBP is buying from Clearview AI
The agreement gives United States Customs and Border Protection access to Clearview AI, which compares photos against “over 60+ billion publicly available images.” Those images come from the internet, where Clearview’s business model has drawn scrutiny because photos are scraped from public websites at scale.
Once collected, the images are converted into biometric templates without the knowledge or consent of the people photographed. That feature is central to why the deal is sensitive: a face image is not just another data point. It can become an identifier that connects a person to other images, records, and relationships.
The contract extends access to Border Patrol’s headquarters intelligence division (INTEL) and the National Targeting Center. These are not casual users. They are units involved in collecting and analyzing information as part of what CBP calls a coordinated effort to “disrupt, degrade, and dismantle” people and networks viewed as security threats.
CBP says its intelligence units use a “variety of sources.” Those sources include commercially available tools and publicly available data, which can be used to identify people and map connections for national security and immigration operations.
Why “tactical targeting” changes the stakes
The contract says Clearview will be used for “tactical targeting” and “strategic counter-network analysis.” Those phrases suggest the software is meant to sit inside the normal workflow of analysts, rather than serve as a narrow tool pulled out only for isolated investigations.
That distinction is important. A face-search tool used rarely has one kind of risk profile. A face-search tool embedded into daily intelligence activity has another, because it can become part of how targets are identified, linked, and prioritized.
The agreement also anticipates that analysts may handle sensitive personal data, including biometric identifiers such as face images. It requires nondisclosure agreements for contractors with access.
At the same time, the source material leaves several operational questions unresolved:
- It does not specify what kinds of photos agents will upload.
- It does not say whether searches may include US citizens.
- It does not explain how long uploaded images or search results will be retained.
- CBP did not immediately respond to questions about integration, authorized image types, or searches involving US citizens.
Those gaps sit at the center of the public concern. The tool’s usefulness depends on matching images and finding possible identities or connections, but the available details do not show the limits around that process.
Scrutiny is growing around DHS face recognition
The Clearview contract comes as the Department of Homeland Security is already under pressure over face recognition in federal enforcement operations. The scrutiny is not limited to border settings. The source article points to large-scale actions in US cities that have swept up US citizens.
Civil liberties groups and lawmakers have questioned whether face-search systems are becoming routine intelligence infrastructure. Their concern is that the technology may no longer be treated as a limited investigative aid, and that safeguards may not have expanded at the same pace as deployment.
Last week, Senator Ed Markey introduced legislation that would bar ICE and CBP from using face recognition technology altogether. The stated concern is that biometric surveillance is being embedded without clear limits, transparency, or public consent.
Clearview AI did not immediately respond to a request for comment.
How the deal fits with DHS systems
Clearview AI also appears in DHS’s recently released artificial intelligence inventory. That entry is linked to a CBP pilot initiated in October 2025 and connects the pilot to CBP’s Traveler Verification System.
The Traveler Verification System conducts face comparisons at ports of entry and other border-related screenings. CBP’s public privacy documentation says that system does not use information from “commercial sources or publicly available data.”
Because of that, the source article says it is more likely, at launch, that Clearview access would be tied instead to CBP’s Automated Targeting System. That system links biometric galleries, watch lists, and enforcement records, including files tied to recent Immigration and Customs Enforcement operations in areas of the US far from any border.
This distinction matters because Clearview’s database is built from commercially gathered and publicly available images. If the technology is routed through targeting and intelligence systems rather than traveler verification, it may function less like a border identity check and more like a broad investigative search layer.
Accuracy limits remain a central issue
Recent testing by the National Institute of Standards and Technology evaluated Clearview AI among other vendors. The testing found that face-search systems can perform well on “high-quality visa-like photos” but struggle in less controlled settings.
Images captured at border crossings that were “not originally intended for automated face recognition” produced error rates that were “much higher, often in excess of 20 percent, even with the more accurate algorithms,” federal scientists say.
NIST also identified a core tradeoff. Face-search systems cannot reduce false matches without increasing the risk that the system fails to recognize the correct person.
For that reason, NIST says agencies may use the software in an “investigative” setting, where it returns a ranked list of candidates for human review rather than one confirmed match. But there is another problem when systems are configured to always return candidates: searches for people who are not already in the database still generate “matches” for review. In those cases, the results will always be 100 percent wrong.
That is the technical backdrop for the CBP deal. The contract points to wider operational use of Clearview AI inside intelligence work, while the public record still leaves basic questions unanswered about scope, retention, affected people, and how uncertain matches will be handled.