us customs And the Border Patrol will spend $225,000 for one year of access to Clearview AI, a facial recognition tool that compares photos to billions of images collected from the internet.
The agreement expands access to the Clearview tool to Border Protection Headquarters Intelligence Division (INTEL) and the National Targeting Center. These units are units that collect and analyze data as part of what CBP calls a coordinated effort to “disrupt, degrade, and dismantle” people and networks that CBP deems security threats.
The contract states that Clearview will provide access to “more than 60 billion public images” and will be used for “tactical targeting” and “strategic counter-network analysis,” indicating that the service is not intended to be dedicated to individual investigations, but to be integrated into analysts’ daily intelligence activities. CBP said its intelligence units use “a variety of intelligence sources,” including commercially available tools and publicly available data, to identify people and map their connections for national security and immigration activities.
The contract assumes that analysts will be handling sensitive personal data, including facial images and other biometric identifiers, and requires non-disclosure agreements for contractors with access. It does not specify what types of photos agents will upload, whether searches will include U.S. citizens, or how long uploaded images and search results will be retained.
The Clearview deal comes as the Department of Homeland Security faces increased scrutiny over how facial recognition is used in federal enforcement operations far beyond the border, including large-scale operations in U.S. cities involving U.S. citizens. Civil liberties groups and lawmakers have questioned whether facial search tools are being deployed as routine intelligence infrastructure rather than limited investigative aid, and whether safeguards are keeping up with the expansion.
Last week, Sen. Ed Markey introduced a bill that would completely ban the use of facial recognition technology by ICE and CBP, citing concerns that biometric surveillance would be incorporated without clear limits, transparency, or public consent.
CBP did not immediately respond to questions about how Clearview will be integrated into its systems, what types of images agents are allowed to upload, and whether searches may include U.S. citizens.
Clearview’s business model has come under scrutiny because it relies on scraping photos from public websites at scale. These images are converted into biometric templates without the knowledge or consent of the people photographed.
Clearview also appears in DHS’ recently released Artificial Intelligence Inventory related to the CBP pilot that began in October 2025. This inventory entry ties the pilot to CBP’s Traveler Authentication System. The system will conduct facial comparisons and other border-related checks at ports of entry.
In its public privacy document, CBP says its traveler verification system does not use information from “commercial sources or publicly available data.” At launch, access to Clearview will likely be tied to CBP’s automated targeting system instead. The system links enforcement records, including biometric galleries, watch lists, and files associated with recent Immigration and Customs Enforcement operations in areas far from U.S. borders.
Clearview AI did not immediately respond to a request for comment.
A recent test by the National Institute of Standards and Technology evaluating Clearview AI with other vendors found that the facial search system worked well with “high-quality visa-like photos,” but poorly in poorly managed settings. Images taken at border crossings that were “not originally intended for automatic facial recognition” had error rates “much higher, often exceeding 20 percent, even when using more accurate algorithms,” federal scientists said.
This test highlighted a major limitation of this technology. NIST found that facial search systems cannot reduce false matches without increasing the risk of not recognizing the correct person.
As a result, NIST says government agencies may operate the software in a “survey” environment, returning a ranked list of candidates for human review rather than a single confirmed match. However, even if the system is configured to always return candidates, searching for someone who does not already exist in the database will generate “matches” for consideration. In these cases, the results are always 100% wrong.
