Immigration in the Age of AI: The Debate Over Biometric Tracking in Enforcement Agencies

A Brief Timeline of DHS AI Directives

In April 2023, the Department of Homeland Security (DHS) issued preliminary Artificial Intelligence (AI) policies to establish a framework for responsible AI implementation in the department’s biometric and technological practices. Policy Statement 139-06 discussed the continuous responsible use of AI, while Directive 026-11 established the DHS’s incorporation of AI directives for facial recognition and capture technology. Subsequently, on September 14, 2023, the DHS announced its official implementation of AI into the department’s immigration protocols. Shortly after, the Biden Administration released the 2024 AI Roadmap, a description of federal AI uses in identifying sector-specific risks, guidance on authentication, and security reviews. 

Increasing AI Implementation: The Debate

AI directives have been extensively deployed in 2024, with the DHS reviewing dozens of cases of AI use across federal agencies. Out of 105 total AI use cases in 2024, 59 were within Customs and Border Protection (CBP), 23 were within Immigration and Customs Enforcement (ICE), 18 were within U.S. Citizenship and Immigration Services (USCIS), and 5 were within other DHS operations. Despite the DHS’s growing trajectory of AI usage, many critics cite potential rights-based abuses with AI-integrated biometric software. Opposing views suggest that this rapid inclusion of AI biometric technologies will boost organizational and procedural efficiency in areas like detainee intake. 

DHS AI Use Cases by Component [All Components]—Yazmin Muñiz, 2025

Arguments in Favor of DHS AI Directives

Many argue that the DHS is taking a crucial step towards efficiency by implementing AI, especially regarding the review of documentation. The I-9 document, used to verify identity and authorization to work legally in the U.S., can take up to 5 federal working days to complete. Supporters claim that AI implementation will allow I-9 auditors to more efficiently sift through documentation, flag discrepancies or potential violations, and verify documents such as passports, green cards, and visas.

Pro-AI arguments also point towards the technology’s potential to facilitate interviewing skills, aiding DHS officers in interrogations, detainee-officer communications, de-escalation, etc. Proponents further highlight AI’s ability to streamline DHS training processes. Training federal employees is often a difficult task—especially during administrative shifts, where laws and policies shift rapidly. Many argue that AI, being repeatable and dynamic, will help make training more efficient and allow it to remain effective amidst changing regulations. 

Proponents have further argued that the DHS’s use of AI will bring improved organization of crucial data, including metadata: the information used to describe and contain other data. With over nine standalone components comprising the DHS, high-risk data is exchanged regularly in high volumes. Likewise, Biden’s 2024 AI Roadmap detailed plans for a robust data framework aimed at utilizing ontological forms of organization to efficiently rearrange data. The framework also includes plans for metadata management, which would improve the organization of properties, data, and sources in an effort to clarify information. 

To achieve all this, the U.S. Coast Guard and DHS collaborated to use the SURVEYOR Integrated Data Environment—a cloud-based system designed to consolidate data and enable data-informed operations. This system also facilitates and streamlines the deployment of new models and ensures data is handled securely. Supporters make the case that without AI directives in the DHS, the streamlining of information and cybersecurity would become difficult to manage, causing efficiency roadblocks in the near future. 

Arguments Against DHS AI Directives

Those against the DHS’s use of AI Directives cite rights-based abuses that can stem from an overreliance on automated operations. For example, ICE typically uses translators to inform migrants of their rights. However, they’ve recently begun working on an automated translation tool with integrated AI to expedite the process. While the government emphasizes its commitment to AI transparency, opponents argue that the human tendency to fall back on automation tools may create an obstacle. With over 59,207 migrants in detention as of September 2025, more than 31 percent of them feel that they cannot communicate effectively, making the need for translation services more pressing. Such high numbers of detainees may lead to piling casework and prompt an increased reliance on AI systems for decision-making.

Similarly, many opponents argue that the potential for faulty pattern recognition may lead to errors in AI’s identification of undocumented migrants. For instance, algorithmic bias can occur when AI models mistake frequency for the root cause–an error that can disproportionately target migrants and people of color (POC). This is evidenced by the DHS’s Homeland Advanced Recognition Technology (HART) system, which compares fingerprints, iris scans, DNA, and more to target certain immigrants for surveillance. Such methods mirror other forms of anti-Black systemic racism, resulting in racial profiling and higher detention rates for POC. Algorithmic bias may thus direct federal operations towards communities of color and unequally affect certain populations, such as Black migrants.

Many also express concerns that AI would allow the DHS to breach civil liberties through surveillance, data retention, and unreasonable searches and seizures. Under the current Trump Administration, the position of Chief AI Officer remains empty following Eric Hysen’s resignation in April 2025. Hysen’s vacancy raises concerns about AI strategy and oversight, with many arguing that the instability is affecting the proper assessment of rights-impacting cases. In 2024, the DHS issued preliminary claims that facial recognition technology isn’t directly rights-impacting. However, CBP stores facial biometric data for up to 12 hours for citizens, and up to 75 years in the IDENT database, creating a strong set of data that could be used for surveillance. Likewise, the upcoming HART system’s gathering of biometric data is often done without a warrant, which breaches the Fourth Amendment’s standards for reasonable search and seizure. 

Conclusion & Future Outlook

As of October 2025, it’s clear that AI will play an increased role in DHS processes. Recent events point to massive software developments, including a recent $30 million contract with Palantir to develop ImmigrationOS—a biometric identification and storage database that draws from public and private data to track undocumented migrants. ImmigrationOS will include AI usage, with those opposed to its development citing the risk of opaque flagging and increased surveillance in workplaces that employ foreign nationals via H-1B or H-2A applications. In the growing race for efficiency, weighing the risks in rights impacts, surveillance, and privatization of AI directives may prove essential for DHS to function ethically.