DHS AI Inventory Update Shows Compliance Gaps, Unexplained Changes

The Department of Homeland Security published an updated artificial intelligence inventory this week that documents risk management practices for high-impact AI systems, but the changes have sparked concerns about transparency and consistency in how the agency evaluates and deploys these tools.

The inventory arrived more than three months after the April 3 deadline set by the Office of Management and Budget. While it includes newly completed risk management sections for high-impact use cases, DHS also reclassified several systems, altered deployment statuses, and retired tools without consistently providing explanations for the changes.

Why it matters

Federal AI inventories are meant to provide public accountability for systems that can significantly affect rights, safety, and legal outcomes. When agencies downgrade risk classifications or remove safeguard documentation without explanation, it undermines the transparency these inventories are designed to create—particularly for systems used in immigration enforcement, border security, and travel screening where individual liberties are at stake.

Inconsistent risk classifications

DHS downgraded several AI systems from "high-impact" to "presumed high-impact but determined not," a classification that exempts them from stringent risk management requirements. Customs and Border Protection made this change for its "Traveler Self-Service Mobile Identity" system, erasing previously documented risk management practices without justification.

The agency did explain its downgrade of the "Consular Consolidated Database," stating that the facial recognition tool used during visa screenings serves only as an advanced search function rather than a principal basis for decisions. However, this inconsistency in providing rationales makes it difficult to understand the criteria DHS applies when reclassifying systems.

One system moved in the opposite direction: TSA's "Answer Engine" screening tool was upgraded to high-impact status. The tool went live on June 11—more than two months after OMB's compliance deadline—and DHS indicated that an appeal process, typically required for high-impact systems, was "not applicable."

Deployment status changes raise questions

Immigration and Customs Enforcement moved its "License Plate Capture and Analysis" tool back to pilot status and removed all risk management information, despite the system becoming operational in September 2025. High-impact pilots are exempt from full compliance requirements, but OMB guidance states minimum practices should be "applied where practicable."

ICE also updated its "Real-Time Language Translation Services" system from pre-deployment to deployed status, but disclosed a June 2025 deployment date—meaning the high-impact tool used in immigration interviews had been operational for a year before the public inventory reflected it.

Retired systems

DHS retired four use cases since January, including two controversial ICE tools. The "Hurricane Score" system, which assigned risk scores to noncitizens in community supervision, had been operational since February 2019. The agency also retired an AI-assisted resume screening tool deployed in January 2026, amid broader scrutiny of ICE's rapid workforce expansion.

TSA's "AskTSA" customer service assistant was removed from the inventory entirely without explanation, despite being deployed since December 2024.

"DHS deserves credit for revisiting its AI inventory rather than just letting it gather dust for a year," Tom Bowman, policy counsel at the Center for Democracy & Technology, told FedScoop. "But the revisions themselves are puzzling and uneven."

The details were first reported by FedScoop.