AI Comes to Immigration Court: A New EOIR Memo Raises More Questions Than Answers
EOIR Acting Director Sirce Owen says AI is "not prohibited" in immigration court, but the memo signals bigger changes ahead with Trump admin encouraging AI use to accelerate government processes.
The Executive Office for Immigration Review (EOIR), the agency that runs the immigration courts, quietly released a policy memo last week that should have everyone who is concerned about the role of AI in the immigration system paying attention.
Although the memo does not contain any revelations on its surface, when taken together with recent research on the use of AI in immigration systems and the White House’s recent announcement encouraging AI adoption across the government, the memo does serve as a starting point for what is sure to be a long and controversial debate over how AI might be weaponized against immigrants facing deportation—especially those seeking asylum.
In this post, I discuss the contents of the memo and connect it to recent research on the use of AI in immigration enforcement systems around the world. Special thanks to Dan Kowalski for bringing this memo to my attention.
Join me in celebrating over 300 Substack posts! This newsletter is only possible because of your support. If you believe in keeping this work free and open to the public, join the 100% of surveyed paid subscribers who said this newsletter was a good steward of their investment today.
The Memo: “Use of Generative Artificial Intelligence in EOIR Proceedings”
Acting EOIR Director Sirce Owen takes a surprisingly procedural and balanced approach in the memo titled "Use of Generative Artificial Intelligence in EOIR Proceedings"—at least compared to her more sensationalistic memos. The memo echoes Chief Justice John Roberts' observation that "AI has great potential to dramatically increase access to key information for lawyers and non-lawyers alike.”
This development is consistent with my previous research on artificial intelligence versus authentic insight, which found considerable uncertainty and anxiety in the legal field about the role AI might play. The rush toward technological solutions often conflicts with the careful, nuanced thinking that complex legal problems require—especially in immigration cases where lives hang in the balance.
Owen establishes that the EOIR has "neither a blanket prohibition on the use of generative AI in its proceedings nor a mandatory disclosure requirement regarding its use"—meaning there's no formal rule stopping anyone from using AI tools in immigration court filings or decisions. Instead, the memo points to existing guidance on professional ethics, thereby subsuming questions of generative AI under existing frameworks. This follows a fairly predictable teleology of how law and policy typically address new technologies: ignore (which they’ve done for a few years), subsume (use existing law to cover novel technology), then, hopefully, adapt (create new laws to handle new technology).
Mirroring concerns from most professors returning to the classroom for the fall, the EOIR memo spends considerable time warning about the risks of hallucinated legal citations and false information generated by AI tools—an entirely reasonable concern, in my view. It makes clear that attorneys who submit AI-generated content with fake case citations or erroneous information will face professional discipline, potentially including sanctions for "knowingly or with reckless disregard offering false evidence." The memo also admonishes that “EOIR adjudicators should also be vigilant to ensure that decisions are not based on hallucinated or inaccurate case citations and information and that attorneys are not submitting pleadings with false information or non-existent legal citations.”
All well and good. But we should not allow the otherwise perfunctory memo to lull us into complacency about how AI is already being leveraged in immigration systems around the world and will likely be increasingly incorporated into institutions here in the United States.
The memo cites the White House’s announcement in April that calls on all parts of government to accelerate the adoption of AI. That memo states: “Agencies are encouraged to harness solutions that bring the best value to taxpayers, increase quality of public services, and enhance government efficiency.” While I am not principally opposed to the government adopting AI or other technologies, we know what this language means when applied to immigration: use AI to reject, deny, and deport faster.
Thus, while the memo does appropriately encourage cautious adoption, we should be clear-eyed about the real implications of AI adoption on due process and substantive outcomes for immigrants facing deportation in court.
Foreshadowing Bigger Changes
As I said, while this memo appears procedural on its surface, it foreshadows significant policy changes down the line and possibly new systems of automated exclusion that we have seen elsewhere in the world.
I have written a lot about the use of smartphone apps like CBP One (and now CBP Home) and SmartLINK, which represent forms of digital capture of migrants’ smartphones. Others, like David Hausman, have written about ICE’s use of risk assessment algorithms to shape decisions about whether immigrants should be detained or released.
These developments reflect a broader trend documented by researchers and policy analysts. A recent report by the American Immigration Council revealed that DHS already operates 105 active AI use cases across major immigration agencies, with 27 classified as "rights-impacting." Legal analysis shows how AI is transforming immigration enforcement and creating new compliance challenges for immigration law practice.
The EOIR memo comes at a time when AI and automated systems are already being integrated into immigration systems across multiple countries. The United States is behind other countries in the adoption of AI. International research paints an even broader picture:
A landmark 2020 study by Citizen Lab at the University of Toronto found extensive use of automated decision-making tools in Canada's refugee system, raising concerns about fairness and transparency.
A recent report from the Council of Europe calls for prohibiting AI tools like automated credibility assessments in asylum cases, while policy research warns that asylum seekers' interactions with AI will serve as a "test case" for emerging AI governance.
Studies of asylum systems have raised particular concerns about AI systems perpetuating bias and creating "black-box" decision-making processes that lack transparency and accountability.
A recent study by Iazzolino (2025) explores how AI-driven migrant screening systems create fundamental trade-offs between efficiency and human rights protections.
Taken together, these studies consistently highlight the tension between efficiency gains and fundamental rights protections that the EOIR memo signals could be coming to US immigration courts.
The Power Dynamics Problem
What we learn from this research—and what concerns me most—is what the memo doesn't say about how power relations within the legal field might shape AI adoption. While the memo doesn't suggest this explicitly, it's not hard to imagine scenarios where immigration judges and asylum adjudicators use AI tools to speed up case rejections—potentially avoiding consequences for irresponsible AI use—while immigration attorneys who use AI face harsher professional consequences or future prohibitions.
Immigration judges are already under enormous pressure to move cases quickly through a system with 3.5 million pending cases. The average judge handles hundreds of cases, often with minimal support staff. As I discussed recently, asylum cases are being denied at record rates and likely at record speeds, too. Adding AI tools that can quickly generate legal research or draft portions of decisions could dramatically accelerate and automate key parts of the adjudication process.
An Early Warning
AI in immigration courts isn't a question of if—it's a question of how and when. This memo represents an early attempt to establish professional norms before the technology becomes ubiquitous, but it also serves as an early warning that the immigration system we know today may look very different in the near future.
The challenge ahead is ensuring that any integration of AI tools serves the cause of equality and justice rather than simply the goal of efficiency. Given the life-or-death stakes of many immigration cases, getting this balance right isn't just a matter of professional best practices—it's also a matter of fundamental fairness. Immigration attorneys, advocates, and anyone who cares about due process should be paying close attention to how this unfolds.
For Further Reading
For those interested in exploring the broader context of AI in immigration systems, here are some recent studies and reports:
Beduschi, A. (2021). International Migration Management in the Age of Artificial Intelligence. Migration Studies, 9(3), 576–596.
Ekpeowoh, S.O. (2024). The Intersection of Immigration Law and Human Rights: Current Issues and Future Trends. Journal of Research and Innovation in Social Science.
Iazzolino, G. (2025). Trading Efficiency for Control: The AI Conundrum in Migration Management. Cosmopolitan Civil Societies: An Interdisciplinary Journal.
Latonero, M., & Kift, P. (2019). On Digital Passages and Borders: Refugees and the New Infrastructure for Movement and Control. Social Media + Society.
Support Public Scholarship
This newsletter is only possible because of your support. If you believe in keeping this work free and open to the public, consider becoming a paid subscriber. You can read more about the mission and focus of this newsletter and learn why, after three years, I finally decided to offer a paid option. If you already support this newsletter financially, thank you.
"We know what this language means when applied to immigration: use AI to reject, deny, and deport faster."
Exactly. There's a huge shortage of immigration judges to deal with the backlog of cases, but there's no indication we'll see any more humans hired, just the addition of more soulless computer programs tasked with making decisions of life and death. We need more "authentic insight" in our immigration system, not less.
What could possibly go wrong with a court system dependent on or even run by artificial intelligence the kind of thing that’s owned by handful of men, and the richest men in the world?