What is HART?
Artificial intelligence (AI) is the newest consideration for the Department of Homeland Security’s (DHS) long-developed Homeland Advanced Recognition Technology (HART) system. HART will harbor the capacity to identify fingerprints, facial features, iris scans, voice patterns, and possibly DNA. In 2015, the DHS announced HART as the successor to the Automated Biographic Identification System (IDENT), which had served as the DHS’s central database since 1994. As of 2026, the IDENT system does not use artificial intelligence, and instead relies on older hardware stored in DHS-owned data centers, which limits how much data the system can handle. In contrast, HART will operate as a cloud-based system, ensuring capacity for hundreds of millions more individuals registered within the system. As a result, HART functions as a central technology matching individuals to law enforcement, immigration, and national security databases.
HART has remained a contentious topic due to its use of AI in wide-scale surveillance. Supporters of HART’s AI inclusion point to its increased capacity for data networking, meaning the ability to share biometric data across many government agencies. They argue this could make criminal and organizational investigations much more efficient. Meanwhile, organizations like the Immigration Defense Project cite concerns over the DHS’s lack of assurance in the accuracy of artificial intelligence systems, with small agencies within DHS and third parties handling sensitive data.
Potential Upsides of HART
In June 2017, JetBlue Airways, in partnership with United States Customs and Border Protection (CBP), tested the first biometric system using facial recognition technology. This partnership led to the implementation of streamlined technology within HART, including an Amazon Web Services (AWS) GovCloud environment. A cloud environment refers to a shared database that major companies use to organize and track data. AWS’s cloud environment will facilitate the expansion and efficiency by allowing longer data retention and faster reporting. With updated capabilities, HART aims to retain user data 75 to 100 years after their date of birth to easily recall information at major check-in points such as airports, border crossings, and detention centers. Same-day biometrics —the collection of identifying markers within a 24-hour period —draws on research suggesting faster transaction times for biometric data. In other words, data processing increases by 25 percent. In October 2025, CBP estimated that HART’s biometric entry-exit system, which operates on faster transaction times, will be feasible within the next three to five years.
HART will work as a basis for other DHS sub-agency projects, such as Immigration and Customs Enforcement (ICE) Repository for Analytics in a Virtualized Environment (RAVEn) platform. In May 2020, RAVEn was launched to investigate human rights violations, including smuggling, trafficking, narcotics, financial, and cyber crimes. RAVEn’s use of AI has been at the forefront of ICE, in combination with HART; both AI-integrated systems will mutually feed biometric data for faster streamlining. HART will operate as a repository of information for RAVEn, allowing access to biometrics and personal data, including addresses, financial information, and even license plate numbers. Cross-examination of non-citizen and citizen data through AI has already seen successful results with CBP Machine Learning Models (MLMs). In 2023, one of CBP’s MLMs flagged a car at San Ysidro, a point of entry along the Mexican border, which contained over 75 kilograms of narcotics.
Potential Downsides of HART
Reservations about HART include the surveillance tactics used to store the data of over 270 million people, including juveniles. HART’s usage extends to subagencies of DHS, including ICE, CBP, and other agencies that have access to any data in the repository. With extensive data, HART’s AI algorithm will likely operate on pattern recognition, which omits important variables surrounding race, socioeconomic status, and gender. In December of 2020, three Black men—Robert Williams, Nijeer Parks, and Michael Oliver—were incorrectly arrested due to faulty facial recognition match results. These false matches often happen because facial recognition systems struggle to accurately identify People of Color, leading to real-world consequences like wrongful arrests. A recent finding by Robert Wolfe and Aylin Caliskan noted that 97 percent of white-identifying individuals were classified as Americans by AI systems, whereas only three percent of Asian-identifying individuals were correctly identified as Americans. This may have real-world consequences if AI is leveraged to enforce immigration regulations within the United States.
Moreover, HART faces limited public oversight and transparency issues, which have decreased trust in the program. DHS has only completed five of the 12 Office of Management and Budget privacy requirements, which are the steps required before launching a system of this scale. Particularly, there was an omission of the individuals whose data will be shared, as well as the partners involved in data sharing. The Office of the Inspector General attributed the omission to a lack of oversight from the Office of Biometric Identity Management (OBIM), mentioning that HART’s protections for personal data are flawed.
Critics also point to a lack of transparency in HART’s funding. In 2018, contractors such as Northrop Grumman received more than $95 million from DHS. In 2020, Peraton, a subsidiary of the private equity firm Veritas, acquired Grumman’s government technology business. Over four billion dollars from various private sources has been invested in HART since it was first announced in 2015. The transition of private partners over the years has led to transparency concerns over data sharing and privacy, especially given the relative lack of privacy regulations for private companies.
Conclusion
AI remains a central part of the development of HART. Under the current Trump administration, responsibility for the project has shifted to the Department of Homeland Security’s Chief Information Officer, signaling a push to speed up development after years of slow progress. DHS argues that HART will improve efficiency at borders, airports, and other security checkpoints by allowing faster and more accurate identity matching. At the same time, HART raises serious concerns about privacy, transparency, and bias. The system’s reliance on AI has already been linked to misidentifications that disproportionately affect People of Color, and DHS has not fully met federal privacy requirements for a system of this scale. Ultimately, the debate around HART represents a difficult balance between greater efficiency for law enforcement and potential risks to civil rights and private data. How DHS addresses these concerns will shape the future role of AI in U.S. biometric systems.