Little-Known Federal Software Can Trigger Revocation of Citizenship – The Intercept
2021/08/27 Leave a comment
Good long read. As AI and automation continue to become more important to manage immigration and other programs, the importance of getting the algorithms and the like becomes more important.
Use of AI in Canada’s visitor visa program provides an example of a measured approach that improves efficiency with appropriate checks.
But as we know from any number of studies, there are consistency and fairness issues with human decision makers as well:
SOFTWARE USED BY the Department of Homeland Security to scan the records of millions of immigrants can automatically flag naturalized Americans to potentially have their citizenship revoked based on secret criteria, according to documents reviewed by The Intercept.
The software, known as ATLAS, takes information from immigrants’ case files and runs it through various federal databases. ATLAS looks for indicators that someone is dangerous or dishonest and is ostensibly designed to detect fraud among people who come into contact with the U.S. immigration system. But advocates for immigrants believe that the real purpose of the computer program is to create a pretext to strip people of citizenship. Whatever the motivation, ATLAS’s intended outcome is ultimately deportation, judging from the documents, which originate within DHS and were obtained by the Open Society Justice Initiative and Muslim Advocates through Freedom of Information Act lawsuits.
ATLAS helps DHS investigate immigrants’ personal relationships and backgrounds, examining biometric information like fingerprints and, in certain circumstances, considering an immigrant’s race, ethnicity, and national origin. It draws information from a variety of unknown sources, plus two that have been criticized as being poorly managed: the FBI’s Terrorist Screening Database, also known as the terrorist watchlist, and the National Crime Information Center. Powered by servers at tech giant Amazon, the system in 2019 alone conducted 16.5 million screenings and flagged more than 120,000 cases of potential fraud or threats to national security and public safety.
Ultimately, humans at DHS are involved in determining how to handle immigrants flagged by ATLAS. But the software threatens to amplify the harm caused by bureaucratic mistakes within the immigration system, mistakes that already drive many denaturalization and deportation cases. “ATLAS should be considered as suspect until it is shown not to generate unfair, arbitrary, and discriminatory results,” said Laura Bingham, a lawyer with the Open Society Justice Initiative. “From what we are able to scrutinize in terms of the end results — like the disparate impact of denaturalization based on national origin — there is ample reason to consider ATLAS a threat to naturalized citizens.”
“From what we are able to scrutinize in terms of the end results … there is ample reason to consider ATLAS a threat to naturalized citizens.”
Some critics believe it’s no accident that ATLAS could go after individual immigrants for flimsy reasons. “The whole point of ATLAS is to screen and investigate so that the government can deny applications or refer for criminal or civil or immigration enforcement,” said Muslim Advocates’ Deborah Choi. “The purpose of the secret rules and predictive analytics and algorithms are to find things to investigate.”
The Department of Homeland Security refuses to disclose to the public how exactly ATLAS works or what rules it uses to determine when an immigrant should be flagged to potentially have their citizenship revoked. This secrecy makes it nearly impossible to tell whether ATLAS is targeting immigrants baselessly or not. The Open Society Justice Initiative this week filed a new FOIA request with DHS and its United States Citizenship and Immigration Services, or USCIS, division seeking details on how the algorithm functions.
The revelations about ATLAS come as policymakers await a review of denaturalization policies that the Biden administration began in February to “ensure that these authorities are not used excessively or inappropriately,” as the White House put it at the time. President Joe Biden came to office promising a more “humane” approach to immigration than former President Donald Trump, who stripped dozens of naturalized Americans of their citizenship. A deadline related to the review came and went in May. Months later, the administration has yet to publish the review or speak publicly about the matter.
ATLAS originates within USCIS, a DHS division with responsibility for granting citizenship and other immigration benefits. USCIS has called the software its “primary background screening system,” but ATLAS appears to be a feature of a larger computer program that helps manage case information on every person in the immigration system: USCIS’s Fraud Detection and National Security Data System, or FDNS-DS. A 2020 DHS assessment of ATLAS’s privacy implications, one of the few public sources of information about ATLAS, shows that when an individual’s information is run through the software — a virtual certainty for any immigrant — ATLAS autonomously scours the databases, including some that contain classified materials.
ATLAS appears to scrutinize not just individual immigrants but also their wider social networks. A 2016 privacy assessment of FDNS-DS said that ATLAS “visually displays linkages or relationships among individuals to assist in identifying non-obvious relationships… with a potential nexus to criminal or terrorist activities.”
Amazon Web Services, the cloud computing division of the large online retailer, was hosting the ATLAS system as of 2020. That arrangement is one of many instances in which Amazon has sold its services to a controversial Homeland Security initiative targeting immigrants. Amazon has faced protests both from the general public and its own employees demanding that the company cease any further anti-immigrant work; the company did not return a request for comment.
USCIS spokesperson Matthew Bourke declined to answer any questions about ATLAS.
Tracking Millions of Immigrants With Potentially Catastrophic Consequences
It’s unknown how many individuals have been denaturalized via ATLAS. But a 2019 USCIS press release gave some sense of the program’s scale, noting that the program that year processed more than 16 million “screenings” and generated 124,000 “automated potential fraud, public safety and national security detections requiring further analysis and manual review by USCIS officers.”
Immigrants come into contact with ATLAS, according to the 2020 privacy assessment, when one “presents him or herself” to the USCIS for some reason, of which there are many; when “new derogatory information is associated with the individual in one or more U.S. Government systems”; or, according to the 2016 privacy document, whenever “FDNS performs an administrative investigation.” This apparently can happen even after an immigration-related decision has been made: Among the FOIA documents shared with The Intercept is a USCIS memo noting that ATLAS is used to detect “fraud patterns in immigration benefit filings … either pre- or post-adjudication,” suggesting that an immigrant could be subjected to algorithmic scrutiny indefinitely after their filing is approved.
Once the system is triggered, ATLAS eventually decides whether to flag the immigrant in question, but it’s unclear exactly how it arrives at that decision. How ATLAS reasons — that is, its decision-making “algorithm” — is secret. And although DHS documents list a handful of data types ATLAS can potentially search, they do not indicate what sorts of personal information ATLAS will churn through to reach its decision.
The 2020 privacy document states vaguely that “ATLAS contains a rules engine that applies pattern-based algorithms to look for indicators of fraud, public safety, and national security concerns,” a process described as “predictive.” It gives little information about these rules but does state that it is permissible to use ATLAS to target immigrants by race and ethnicity in “exceptional instances,” a term left glaringly undefined. The document claims that USCIS protects immigrants from discrimination by “limiting the consideration of an individual’s simple connection to a particular country, by birth or citizenship, as a screening criterion, unless such consideration is based on an assessment of intelligence and risk and in which alternatives do not meet security needs.” Caveats aside, the point is clear: ATLAS could be used to target certain ethnic groups or nationalities in “exceptional circumstances” or should DHS deem it a “security need.” Appealing to murky notions of “national security” and “fraud” is a long-standing tactic of the post-9/11 homeland security apparatus, and one that has historically permitted the state to justify efforts to harass or target marginalized communities in the U.S. under the auspices of public safety.
If ATLAS produces a negative review, the next steps can lead to denaturalization, and a 2019 flowchart included in the FOIA documents provided to The Intercept illustrates how: When ATLAS finds something derogatory according to its secret list of rules, the software sends out a “System Generated Notification,” which is then “triaged” and forwarded directly to FDNS-DS if potentially “actionable.” From there, FDNS determines whether the notification constitutes a “possible criminal denaturalization referral,” and, if so, will “refer to ICE for criminal denaturalization action.” All told, going from an ATLAS notification to criminal denaturalization proceedings takes only four steps on the flowchart.

An internal USCIS document shows an ATLAS scan as the first step in identifying cases for denaturalization.
Document: FOIA
A USCIS spreadsheet summarizing the System Generated Notifications created in 2020, also obtained via FOIA litigation, cites 12 different categories of ATLAS alert. Though the meaning of these codes is unclear, the spreadsheet references notifications relating to “DACA,” presumably the Deferred Action for Childhood Arrivals policy that protects some undocumented immigrants from deportation; “DOD,” possibly referring to the Department of Defense; and two different “NS,” or national security, categories whose full names were redacted. Most of the notifications created in 2020 were in the “multiple identities” category, which refers to immigrants deliberately using false aliases.
Legal scholars and technologists have widely criticized attempts to use software to predict national security threats, arguing that terrorism is so statistically rare as to be impossible to foresee by drawing “patterns” from a person’s biography. “Because the rules or factors underlying ATLAS’s screening functionality are unknown, there is no way to assess whether ATLAS is disproportionately flagging certain communities,” Choi of Muslim Advocates told The Intercept. “In fact, the Privacy Impact Assessment for ATLAS states that under certain circumstances, an individual’s country of birth or citizenship could be a screening criterion. As was the case in Operation Janus” — a DHS program that involved a review of past naturalization cases of people from “special interest countries” — “any rule based on country of origin is likely to target individuals from Muslim-majority countries.”
The 2020 privacy document does little to dispel worries that ATLAS is making potentially life-ruining decisions on the basis of bad data. The document states that ATLAS’s output is subject to manual review by the agents who use it; it also notes that the accuracy of ATLAS’s input is taken as a given: “USCIS presumes the information submitted is accurate. … ATLAS relies on the accuracy of the information as it is collected from the immigration requestor and from the other government source systems. As such, the accuracy of the information in ATLAS is equivalent to the accuracy of the source information at the point in time when it is collected by ATLAS.” The document further notes that “ATLAS does not employ any mechanisms that allow individuals to amend erroneous information” and suggests that individuals directly contact the offices maintaining the various databases ATLAS uses if they wish to correct an error. The notion that someone struggling to navigate the U.S. immigration system would have the wherewithal to personally negotiate a correction of the FBI Terrorist Screening Database, or have an opportunity to learn of such an error to begin with, is questionable.
An Opportunity To Stop the Denaturalization Wave
The U.S. government’s use of denaturalization has varied widely over the last century. In the early to mid-1900s, the federal government pursued denaturalization for political, racist, and sexist reasons, even going after U.S.-born citizens. That changed after a 1967 U.S. Supreme Court decision vastly narrowed the potential uses of denaturalization. For nearly five decades afterward, the government brought denaturalization cases only sparingly, usually against accused war criminals and Nazis — up until the Trump presidency.
In September 2017, the Department of Justice announced its intent to denaturalize three men it accused of lying about their immigration histories on their applications for citizenship. It was a loud proclamation of a new front in the Trump administration’s war on immigrants that would lead to nearly double the number of denaturalization cases filed during two years as compared to the number of cases filed from 2004 to 2016, according to a New York Times Magazine investigation.
The infrastructure that helped the Trump Justice Department identify its first targets for denaturalization was years in the making. Under Operation Janus — an initiative that began at the end of George W. Bush’s presidency and continued under former President Barack Obama — the Department of Homeland Security began to digitize fingerprint data for about 315,000 people whose information was missing from a central database, ultimately identifying 1,029 people who had been naturalized after receiving final orders of deportation under another identity. According to a 2016 report from the DHS Office of Inspector General, U.S. Immigration and Customs Enforcement had begun the process of investigating some of those cases to decide whether the individuals should be denaturalized.
“But the Obama administration proceeded with caution, instructing officials only to denaturalize those who appeared to pose a danger to the United States,” writes law professor Amanda Frost in her recent book, “You Are Not American: Citizenship Stripping from Dred Scott to the Dreamers.” “After the Trump administration took over, however, the program grew exponentially.”
In early 2018, the Justice Department wrote in a press release that USCIS “has stated its intention to refer approximately an additional 1,600 for prosecution,” and later that year, USCIS announced the creation of a new office focused on denaturalization. (Asked about the status of that office, Bourke, the USCIS spokesperson, said that once the administration’s review of denaturalization policies is complete, “USCIS staffing will be adjusted accordingly to meet the needs of the agency.”) Ahead of the 2019 and 2020 fiscal years, the Department of Homeland Security asked for $207.6 million to fund, among other things, investigations into hundreds of additional leads under Operation Janus, as well as a review of another 700,000 immigrant files under Operation Second Look, a related program. In early 2020, the Justice Department created a new office to investigate “terrorists, war criminals, sex offenders, and other fraudsters who illegally obtained naturalization” for denaturalization.
Source: Little-Known Federal Software Can Trigger Revocation of Citizenship – The Intercept