U.S. Citizenship and Immigration Services pressed to unveil how it uses AI in asylum cases

Via Law360
By Britain Eakin

Refugees International fired off a lawsuit in D.C. federal court on Friday to pry loose records from U.S. Citizenship and Immigration Services about its use of artificial intelligence to process asylum applications, saying the agency has stonewalled its request for nearly two years.

The nonprofit humanitarian organization says the public knows very little about the agency’s implementation of an analytical tool called Pangaea Text, which USCIS reportedly uses to flag fraud, national security and public safety threats, according to the complaint.

Among the details Refugees International is hoping to unearth through its Freedom of Information Act lawsuit is how the program works, its effects on asylum applications, and how the AI tool is trained. Additionally, the organization wants to better understand how USCIS mitigates bias and ensures the tool is accurate.

“The absence of this information risks jeopardizing access to asylum by vulnerable populations and, consequently, impairs the ability of the United States to provide protection through asylum to vulnerable individuals consistent with U.S. law and policy,” Refugees International said.

The organization lodged its FOIA request with USCIS on Dec. 24, 2022, but said the agency has yet to provide any response at all, spurring the lawsuit to force its hand. The nonprofit sought USCIS guidelines on the use of Pangaea Text, along with memoranda, training materials and samples of the program’s output.

Refugees International also wants the agency to turn over data illuminating how many cases are screened with the tool, how many are flagged for fraud or security concerns along with data illustrating the outcomes. It has also asked the agency to say whether asylum-seekers have an opportunity to respond when their cases are flagged by the AI tool, and an accounting of those instances if so.

According to the organization, irresponsible use of AI to assess asylum applications could “intensify existing inequities and introduce new types of harmful discrimination.” For example, it could unfairly flag applications from non-English speaking applicants and those without attorneys who fill out their immigration paperwork themselves, according to Yael Schacher, director of Refugees International for the Americas and Europe.

“Do asylum officers give them a chance to explain this when ATA flags their application as fraudulent?” she said in a statement. “DHS has said it is being careful with its use of AI but has not been transparent about its use of a tool that may be determining whether someone is denied refuge and deported.”

Without more information, the public can’t assess the extent to which USCIS is responding to those concerns in its deployment of the technology, the complaint said.

USCIS did not immediately return a request for comment Friday.

Refugees International is represented by David Bitkower, Ian C. Moss and Lawrence W. McMahon of Jenner & Block LLP, and Sabrineh Ardalan, Martha Ball and Jessenia Class of the Harvard Immigration and Refugee Clinic.

Counsel information for USCIS was not immediately available.

The case is Refugees International v. U.S. Citizenship and Immigration Services, case number 24-cv-3559, in the U.S. District Court for the District of Columbia.