U.S. Customs and Border Protection is a sub-agency of DHS and, by the count in the department's own inventory, one of its largest deployers of artificial intelligence. Its systems fall into three broad functions: biometric identification of travellers, detection and surveillance between and at ports of entry, and risk scoring of cargo and trade entities.
The DHS use-case inventory
DHS publishes a per-component AI use-case inventory. The CBP page, dated July 15, 2026, is the primary public record of what the agency has fielded and at what stage. A structured read of that page returned 69 CBP entries, 43 of them marked "Deployed" (Source: dhs.gov). Among the entries listed:
| Use case | ID | Status | Function |
|---|---|---|---|
| Autonomous Surveillance Tower (AST) | DHS-35 | Deployed | Border Patrol surveillance between ports of entry |
| Automated Item of Interest Detection | DHS-37 | Deployed | Alerting on items of interest in monitored field imaging |
| Vessel Detection | DHS-38 | Deployed | Maritime detection and sensor-network analytics |
| Aircraft Landing Location Predictor (KESTREL) | DHS-65 | Deployed | Predicts likely aircraft landing sites from historical flight paths |
| Empty Container Detection Model | DHS-68 | Deployed | Identifies and tracks empty containers in cargo shipments |
| Commodity Detection Model | DHS-69 | Deployed | Object detection on X-ray images to predict commodity codes |
| Traveler Entity Resolution | DHS-80 | Deployed | Identity verification for travellers |
| Passport Anomaly Model | DHS-81 | Deployed | Flags discrepancies in passport information |
| Agriculture Commodity Model (AGC) | DHS-86 | Deployed | Risk-based inspection sampling for agricultural imports |
| Cargo Classification Tool | DHS-94 | Deployed | Text analysis of product descriptions |
| Trade Entity Risk Model | DHS-95 | Deployed | Supervised models scoring trade-entity risk |
| Advanced Trade Analytics Program (ATAP) | DHS-101 | Deployed | Aggregates and analyses historical and current trade data |
| Non-Intrusive Inspection 3D Imaging Tool | DHS-163 | Pilot | 3D imaging for border and transportation security |
| Public Information Compilation (Dataminr) | DHS-183 | Deployed | Compiles publicly available information into threat alerts |
| Babel | DHS-185 | Deployed | Compiles social-media and open-source information on travellers flagged for further screening |
| Airship Outpost | DHS-188 | Deployed | Identifies and documents cross-border conveyances |
| Unified Processing / Mobile Intake | DHS-398 | Deployed | Border-security processing |
| Facial Recognition for National Security and TCOs | DHS-2729 | Pilot | Border Patrol identification of unidentified individuals |
| Mobile Fortify | DHS-2731 | Deployed | Field application for verifying and identifying subjects of interest |
Traveller identity verification appears as a family of separate entries — officer-initiated, traveller-initiated, third-party, and vetting — each described in the inventory as comparative facial recognition that "does not enter or retrieve data." CBP Link (DHS-2619) is the Traveler Verification System integrated into CBP One, comparing live or uploaded images against CBP's database for real-time identity verification, and Traffic Jam (DHS-2570) applies facial recognition, AI and geospatial analysis to human-trafficking data (Source: dhs.gov).
Trade enforcement
On October 31, 2025, CBP signed a two-year contract with Altana to provide AI-enabled compliance screening. The agency described the screening as "critical" for enforcing Section 232 national security tariffs on steel, aluminum, and other products.
The inventory's trade entries — ATAP, the Trade Entity Risk Model, the Cargo Classification Tool, the commodity and empty-container detection models, and Entity Resolution (DHS-24), which CBP describes as aggregating and analysing global trade data across multiple languages for network analysis — are the in-house counterpart to that contracted screening (Source: dhs.gov).
Border biometric identification
CBP uses facial recognition for border-crossing identification. Reporting by MIT Technology Review on August 14, 2024, under the headline "The US wants to use facial recognition to identify migrant children as they age," described the use of AI to track the biometric aging of unaccompanied minors over time.
Google began rolling out machine-learning-based age estimation in the United States on July 30, 2025, a technology adjacent to those used in CBP border-screening contexts.
Mobile Fortify, listed as deployed, moves identification into the field: DHS describes the application as empowering agents "to quickly verify and identify subjects of interest," and it is deployed at both CBP (DHS-2731) and ICE (DHS-2577) (Source: dhs.gov; brennancenter.org). Wired reported in February 2026 that the CBP and ICE face-recognition application cannot in fact verify identity as such, and 404 Media reported in January 2026 that the ICE application misidentified the same woman twice (Source: wired.com; brennancenter.org).
Procurement
Beyond Altana, two procurement lines are documented. CBP contracted with Clearview AI on a one-year base contract giving the agency access to a database of more than 60 billion publicly available images, which FedScoop reported as supporting "tactical targeting" and "counter-network analysis"; Biometric Update described the arrangement as formally integrating Clearview's platform into CBP intelligence and targeting operations (Source: fedscoop.com; biometricupdate.com).
The Brennan Center's tracking of DHS surveillance contracts puts department-wide facial-recognition expenditure from 2021 to 2025 at at least $47.188 million, of which roughly $40 million went to NEC Corporation through the Office of Biometric Identity Management and roughly $7.188 million to Clearview AI across ICE and CBP. Those figures are DHS-wide rather than CBP-specific (Source: brennancenter.org).
Oversight and controversy
Civil-liberties concerns have been raised around the biometric tracking of children and immigrants. Accuracy and bias have also been raised in connection with facial recognition for non-white and non-adult faces (see AI and Surveillance, Algorithmic Accountability and Bias Audits).
The Brennan Center reports that field agents have gained access to real-time facial identification and that the capability has, on reporting it cites, been turned on protesters and observers; it also collects the governing DHS documents, including DHS Directive 026-11 on the use of face recognition and face capture technologies, issued September 2023 and since removed from the DHS website (Source: brennancenter.org). FedScoop reported that lawmakers were moving to constrain the agency's use of these tools even as DHS expanded them (Source: fedscoop.com).
Reporting by 404 Media, public by September 8, 2026, describes Border Patrol operating predictive policing units — indexed by the Marshall Project as Predictive Intelligence Targeting Teams — that analyse Americans' financial activity and other data and pass the resulting intelligence to local police, who stop people not suspected of any specific crime (Source: 404media.co; themarshallproject.org). The article is paywalled and only its opening paragraph was retrieved; the units' inputs, models, scale and governing authority are not established from it. See Predictive Policing.
Open questions
- Whether the Altana contract was renewed or extended past its two-year term in October 2027 is not established here.
- The DHS inventory records deployment status but not model provenance, evaluation results, or error rates for any CBP use case.
- Whether the Predictive Intelligence Targeting Teams appear in the DHS AI use-case inventory, and under what name. The inventory entries recorded above do not obviously correspond to them.
Relationships
- parent: DHS — Department of Homeland Security (AI Deployer)
- related: ICE — Immigration and Customs Enforcement (AI Deployer) — shares the Mobile Fortify application and the Clearview contracting line.
- deployed-by: Clearview AI, Altana AI, NEC Corporation, Dataminr.
- related: AI and Surveillance, Algorithmic Accountability and Bias Audits, Federal AI Adoption — Patterns and Tensions.
- instance-of: Federal AI Compliance Landscape — a component subject to the federal high-impact-use framework.