The Government Accountability Office (GAO) found the Department of Homeland Security’s (DHS) January 2026 artificial intelligence inventory contained required information, but the records provide little detail about the performance of immigration enforcement tools.
GAO’s February finding concerned the information in the filing. It did not establish whether individual systems produce accurate results or what effects they have in practice.
Federal law requires agencies to maintain inventories of current and planned AI uses and make them public when practicable. The Office of Management and Budget (OMB) requires agencies to publish an inventory at least annually and encourages more frequent updates. Its guidance defines high-impact AI as a system whose output serves as a “principal basis” for a decision or action with a legal, material, binding or significant effect on rights or safety. Such systems face safeguards including testing before deployment, impact assessments, monitoring and human oversight.
In a June 12 update to its 2025 inventory, DHS lists Mobile Fortify as a deployed, high-impact tool used by U.S. Immigration and Customs Enforcement (ICE), with an operational date of May 20, 2025. DHS says the app sends facial images, fingerprints and identity-document photos to matching systems operated by U.S. Customs and Border Protection (CBP). ICE’s entry says the agency does not own, train, test or evaluate those models.
The inventory answers a question about monitoring for privacy, civil-rights or civil-liberties harms with: “Yes, sufficient monitoring protocols have been established.” It gives no results from those protocols or a performance rate.
DHS places Enhanced Lead Identification and Targeting (ELITE) in a different category: it is presumed high-impact but was determined not to meet that threshold. DHS says ELITE extracts addresses from records such as rap sheets and warrants, then officers review and validate leads before deciding what action to take. The classification matters because OMB’s high-impact designation determines which safeguards apply.
Rachel Levinson-Waldman and Eoghan Stafford argued in an Oct. 21 analysis for Just Security that the inventory does not give the public enough detail to assess the risks and effects of specific systems. Their concern addresses a different question from GAO’s finding about whether required information was present.
An earlier DHS inspector general audit found 66 AI use cases on the department’s internal inventory that were excluded from public reporting without formal justification. OMB’s 2025 guidance retains impact assessments that include protected-class data and plans to address negative effects, but it does not repeat the 2024 memo’s explicit requirement to assess performance disparities across demographic groups.
DHS cites faster identification and more efficient address extraction as benefits; people whose identities or locations are assessed may bear the consequences of inaccurate information. OMB requires public inventories at least annually and encourages agencies to update them more often.
