DHS Expands AI Use in Immigration Enforcement by 36% in One Year

The Department of Homeland Security has dramatically accelerated its deployment of artificial intelligence in immigration operations, adding 58 new AI systems in just six months between July 2025 and January 2026, according to a new analysis of the agency's updated AI Use Case Inventory.

Active immigration-related AI use cases increased 36% between December 2024 and January 2026. Nearly 90% of the new systems came from Immigration and Customs Enforcement and Customs and Border Protection, with ICE alone adding 28 new AI programs. Most support investigations, surveillance, and targeting operations rather than benefits processing or customer service.

The pace of deployment is notable: 27 of the 58 new systems were already operational when reported, with three more in pilot programs. Only a minority remain in early development stages.

Generative AI Moves to Operational Use

Half of the newly reported systems—29 out of 58—use generative AI, marking a significant shift from last year when predictive analytics and computer vision dominated the discussion. DHS is now using generative AI for document summaries, semantic search, investigative support, translation, and code generation across daily operations.

This technology creates new content rather than simply analyzing existing data, generating summaries and analyses that may influence how personnel interpret cases. ICE's Enhanced Lead Identification and Targeting system, developed with Palantir, uses generative AI to organize investigative leads. Another ICE program analyzes publicly available online information to identify potential targets. A third system searches years of investigative reports and automatically generates summaries.

CBP's RAPTOR system combines computer vision, radar, and infrared imagery to detect and classify border activity.

The High-Impact Classification Question

DHS designated 15 new systems as high-impact AI under Office of Management and Budget guidance, which requires additional oversight and testing for systems that significantly affect rights, government services access, or safety. However, the agency classified another 14 systems as "presumed high-impact but determined not high-impact."

In nearly every case, DHS concluded these systems don't qualify for heightened oversight because they support rather than serve as the principal basis for decisions—a human remains in the decision loop. This classification matters because it determines which federal oversight requirements apply.

The distinction raises questions about effective human review as generative AI becomes more integrated into enforcement operations. Research shows people don't always detect AI-generated errors or hallucinations. ICE's use of facial recognition app Mobile Fortify has already led to wrongful detentions of both immigrants and protesters through misidentification.

Why it matters

As DHS rapidly deploys AI systems that summarize evidence, identify targets, and support detention decisions, the effectiveness of human oversight becomes critical—yet most systems avoid the stricter review requirements that come with high-impact classification. The agency's determination that a human in the loop is sufficient protection assumes reviewers have adequate time, training, and access to underlying data to catch AI errors, an assumption that deserves scrutiny as these tools become embedded in enforcement operations affecting fundamental rights.

Transparency Gaps Remain

The inventory provides basic transparency but leaves significant gaps. Many systems include only brief or vague public descriptions, vendor information is often incomplete, and details about testing, bias assessments, accuracy metrics, and real-world outcomes remain inconsistent.

These details were first reported by the American Immigration Council in an analysis by Steven Hubbard published July 14, 2026.