Unexplained USCIS requests suggest DHS using AI to flag visa applicants

In an unexpected and deeply concerning move, the U.S. Citizenship and Immigration Services (USCIS) has begun issuing Requests for Evidence (RFEs) in H-1B and employment-based immigrant petitions that ask the petitioner for biometric data and detailed residential histories.

This sudden and unprecedented procedural shift has triggered alarms throughout the immigration law community, with attorneys warning that these new RFEs not only deviate from longstanding procedures, but that they also are indicative of a deeper transformation in the role and behavior of USCIS under the Trump administration. The move does come on the heels of an aggressive expansion of AI into high-risk areas of federal surveillance and enforcement that lacks oversight, transparency, and accountability, as Biometric Update reported this week.

Traditionally, RFEs in H-1B petitions serve as a targeted, regulatory function by requesting clarification or additional documentation related to an applicant’s eligibility, such as whether the position qualifies as a specialty occupation, the beneficiary possesses the requisite educational degree, or whether there is a valid employer-employee relationship.

Recent USCIS RFEs though are instead demanding personally invasive information with no clear bearing on these core adjudicatory standards. Immigration attorneys say USCIS has requested for up to five years of residential address history and documentation of physical presence such as lease agreements or utility bills. In other cases, the RFEs appear to be seeking biometric identifiers such as fingerprints and facial photos, despite no biometric appointment notices having been issued through the formal USCIS Application Support Center (ASC) process.

“This is highly unusual because biometrics are not typically required for these case types,” Vic Goel of Goel & Anderson told Forbes. “The RFEs also fail to explain the nature of the adverse information, leaving employers and attorneys in the dark. It appears that [the Department of Homeland Security (DHS)] may be using AI tools to flag individuals based on undisclosed data, possibly from social media or other government databases,” Goel said.

More alarming to immigration experts is that these requests often cite vague claims of “potentially adverse information” as justification for an RFE. Yet, they fail to disclose the nature of that information or how it relates to statutory eligibility criteria. For example, in one such RFE, a USCIS adjudicator wrote, “We have encountered potentially adverse information related to the beneficiary. To continue processing your application or petition, we require an updated address for the beneficiary so that we may collect biometric data.” However, nowhere in the letter is the nature of the adverse information explained.

This kind of ambiguity strikes at the heart of due process. Under 8 CFR §103.2(b)(16)(i), USCIS is required to disclose derogatory information to a petitioner if it serves as the basis for an adverse decision. The agency’s failure to meet that requirement has left attorneys questioning whether USCIS is using RFEs not for adjudication, but rather for pretextual data gathering, possibly as part of enforcement priorities that go beyond the USCIS’s traditional benefits-focused scope.

Kevin Miner of Fragomen similarly noted that these new RFEs “do not ask substantive questions” about the beneficiary’s qualifications or job duties but rather focus solely on vague adverse information and biometric collection. In his view, this is an unprecedented move away from the well-established practice of communicating through documentary exchange.

“It’s very unusual for USCIS to use RFEs in this way, and the agency hasn’t issued any guidance about this new process,” Miner said.

The backdrop for this procedural shift is the Trump administration’s expansive approach to immigration surveillance. Trump executive orders have given immigration authorities broad discretion to use digital tools, data integration systems, and AI to identify, monitor, and remove “inadmissible individuals,” including those who have lawful status.

Under the administration’s new policy environment, the boundaries between USCIS’s role as a benefits adjudicator and Immigration and Customs Enforcement’s (ICE) enforcement mandate appear to be eroding. If USCIS is now acting as a data collection proxy for ICE, it would mark a significant and troubling shift in the agency’s function.

The lack of transparency makes it difficult to gauge the full scope of the policy change, as USCIS has not issued public notices or updated its Policy Manual to reflect any change in biometric collection procedures for employment-based cases. Nor has there been any publication in the Federal Register announcing new rulemaking. This stealth rollout has prompted some immigration attorneys to file Freedom of Information Act requests to uncover whether internal USCIS memos or interagency directives are behind the sudden expansion of RFEs.

While not publicly confirmed in detail, there is evidence that USCIS has explored automated tools for fraud detection and form processing – tools that could potentially influence RFE’s by flagging anomalies or inconsistencies that a human adjudicator might otherwise deem immaterial. Critics worry that overreliance on automated analysis, especially in a legal process as nuanced as immigration adjudication, could lead to rigid interpretations of eligibility criteria and reduce the flexibility that immigration officers have traditionally exercised in evaluating complex cases.

The consequences of the new policy ambiguities are severe. Without knowing the source or purpose of the new data requests, petitioners and their attorneys are unable to mount meaningful responses. In some cases, attorneys have advised their clients not to comply with the RFEs until USCIS has fully explained the basis of the adverse information. Others have opted to legally challenge the requests, citing the agency’s obligation to disclose all material evidence that could affect adjudication.

“USCIS is openly bragging on social media about arresting persons who came into offices for interviews or other normally routine processes, which makes it difficult to advocate that one unwittingly subject themselves to such action,” Goel said.

There also is growing concern that these RFEs may be driven in part by automated flagging systems powered by AI. Goel said DHS could be using AI to detect anomalies or draw inferences from sources like social media or commercial data brokers without disclosing the methodology or affording applicants an opportunity to rebut potentially flawed or biased information. If so, this would align with a broader trend across federal agencies to incorporate AI-enhanced surveillance technology into immigration operations while bypassing traditional oversight mechanisms.

Biometric Update has reported that no federal department embodies the convergence of AI, surveillance, and civil liberties violations more clearly than DHS. A recent analysis of DHS’s AI inventory revealed a vast and obscure digital arsenal that is hidden in bureaucratic shadows. The report highlights nearly 200 AI applications, many previously unknown to the public, that span visa vetting, border enforcement, deportation logistics, and biometric identification.

These and the USCIS’s latest actions coincide with a broader crackdown on lawful immigrants across visa categories. The Trump administration has already drawn criticism for deporting international students based on minor administrative issues or online political activity.

As of mid-April, more than 1,550 international students and recent graduates had their visa statuses altered or revoked according to reporting by Inside Higher Education. Employers and universities fear that heightened scrutiny of H-1B applicants is part of a larger effort to suppress legal immigration under the guise of national security and fraud prevention.

The H-1B visa program is already constrained, with only 85,000 visas available annually and tens of thousands of applicants rejected each year. According to the National Foundation for American Policy, only 20 percent of new H-1B applications are approved annually, while some estimates suggest that “a European teenager is four times more likely to get a seasonal visa to work in an amusement park than a graduate student is to get H-1B status to work in AI.” Similarly, the per-country limit on green cards results in decades-long wait times for highly skilled professionals from countries like India and China.

The new wave of biometric RFEs could exacerbate these challenges by adding new barriers to an already rigid system. The opacity and irregularity of these requests undermine predictability, which is essential for employers making hiring decisions and foreign professionals planning their futures in the U.S. If companies see the U.S. immigration process as arbitrarily punitive or functionally surveillant, they may shift investment and talent recruitment elsewhere.

Meanwhile, the legal architecture governing RFEs is being tested. Under 8 CFR §103.2(b)(8), USCIS is authorized to issue an RFE when initial evidence is insufficient. However, the regulation also mandates that the RFE clearly articulate the missing eligibility requirement and the types of evidence that could satisfy it. That standard appears to be undermined by the new RFEs, which lack specificity and instead raise vague allusions to adverse information.

Historically, RFEs in both the H-1B and I-140 employment-based immigrant visa processes have been used to verify claims related to job eligibility, academic qualifications, wage levels, and employer credibility. During the Trump administration’s first term, RFEs surged dramatically, with issuance rates jumping from 21.4 percent in fiscal year 2016 to nearly 60 percent in 2018.

This spike was largely driven by Trump’s April 2017 “Buy American, Hire American” executive order and internal agency guidance that encouraged adjudicators to deny petitions that lacked narrowly tailored evidence. That trend was reversed under the Biden administration in January 2021 with Executive Order 14005, which sought to restore “fair and consistent” adjudications.

In June 2021, USCIS updated its RFE/Notice of Intent to Deny (NOID) policy to reinstate a 2013 memo encouraging immigration officers to issue RFEs or NOIDs rather than deny cases outright – a reversal of 2018 Trump-era guidance that allowed denials without RFEs in many circumstances.

Today, however, it appears that those gains may be unraveling. Immigration attorneys remain in a “wait and see” posture, uncertain whether the new RFEs mark an isolated deviation or the beginning of a broader campaign to reengineer immigration vetting through invasive data collection and murky AI tools.

Miner cautioned that the requests are still too recent to draw firm conclusions but agreed that the stakes are high. “We haven’t seen these progress far enough to understand the purpose of the requests,” he said, noting, however, that “the absence of transparency is deeply concerning.”

Article Topics

biometric identifiers  |  biometrics  |  border security  |  data collection  |  DHS  |  immigration  |  U.S. Government  |  USCIS  |  visas