CBP Seeks AI/ML Tools to Speed Border Image Analysis

The U.S. Customs and Border Protection (CBP) wants industry input on artificial intelligence and machine learning (AI/ML) tools to support its evaluation of images taken at the border using the agency’s non-intrusive inspection (NII) systems. 

In a request for information (RFI) posted to SAM.gov on April 30, CBP said it is aiming to “reduce the average NII examination processing time” while increasing the number of scanned conveyances. The agency said the goal is to improve security outcomes and reduce officer workload without disrupting commerce. 

The agency’s Office of Field Operations, through its NII Program, is focusing specifically on large-scale imaging systems that generate high volumes of X-ray image data and metadata from privately owned and commercial vehicles. These systems are already deployed across airports, seaports, land border crossings, and other entry points, and are expanding in scale and throughput, CBP said. 

CBP said the growing volume of scan data is outpacing human analysis capacity, creating demand for AI/ML tools that can assist officers with anomaly detection, threat identification, and decision-making.  

The agency emphasized that “frontline personnel remain the most critical asset,” with AI/ML serving as a supporting tool in a human-in-the-loop model. 

Through the RFI, CBP said it is seeking modular algorithmic solutions – rather than full data platforms – that can integrate into existing CBP-owned systems and architectures. The agency said it is not looking for proprietary data pipelines or case management systems, but instead for components that can operate within its current environment. 

Key capability areas outlined in the RFI include anomaly detection, conveyance modification detection, manifest validation, and identification of contraband such as narcotics, weapons, and currency.  

CBP is also interested in tools that can segment images, highlight areas of interest, and assign risk scores to support officer review. 

Industry responses are due by May 30.