DOJ Guidance on Generative AI Use in EOIR Proceedings

OOD PM 25-40 Effective: August 8, 2025 To: All of EOIR From: Sirce E. Owen, Acting Director Date: August 8, 2025 USE OF GENERATIVE ARTIFICIAL INTELLIGENCE IN EOIR PROCEEDINGS PURPOSE: Provide general guidance on the use of generative artificial intelligence (AI) in EOIR proceedings OWNER: Office of the Director AUTHORITY: 8 C.F.R. § 1003.0(b) CANCELLATION: None The use of generative AI has been one of the most significant—and potentially transformative— developments in the legal profession in recent years. As with many technological advancements, however, the increased use of generative AI offers both significant benefits and risks. As Chief Justice John Roberts noted at the end of 2023, “AI obviously has great potential to dramatically increase access to key information for lawyers and non-lawyers alike. . . . But any use of AI requires caution and humility.” 2023 Year-End Report on the Federal Judiciary, U.S. Supreme Court, https://www.supremecourt.gov/publicinfo/year-end/2023year-endreport.pdf. EOIR, too, should approach the use of generative AI with an open mind due to its potential benefits, but also with caution due to its potential risks. Perhaps the greatest potential risk, as demonstrated by multiple media reports in recent years, is the use of hallucinated legal citations or arguments generated by AI and subsequently filed with a court, which causes significant damage to the legal system and may lead to possible sanctions in addition to reputational harm. As the largest administrative court system by case volume in the The use of hallucinated citations is not necessarily limited to pleadings filed by attorneys. Recently, two separate federal district court judges withdrew opinions after attorneys raised questions regarding the accuracy of record citations and information in those opinions. Reuters, Mike Scarcella, Two US judges withdraw rulings after attorneys question accuracy (Jul. 29, 2025), https://www.reuters.com/legal/government/two-us-judges-withdraw-rulings-after- attorneys-question-accuracy-2025-07-29/. Although neither judge confirmed whether generative AI was responsible for the errors, the decisions bore indicia consistent with research or drafting done by generative AI. Although EOIR prohibits “the unauthorized use of AI services. . .on [Department of Justice] Government Furnished Equipment,” EOIR Office of Information Technology, Notification: Updated - Unauthorized use of AI services on GFEs (May 22, 2025), the broader policy of the federal government is “to accelerate the Federal use of AI by focusing on three key priorities: innovation, governance, and public trust.” Office of Management and Budget Memorandum M-25-21, Accelerating Federal Use of AI through Innovation, Governance, and Public Trust, at 2 (Apr. 3, 2025), SIRCE OWENDigitally signed by SIRCE OWEN Date: 2025.08.08 13:41:45 -04'00'

federal government, EOIR is particularly susceptible to the improper or problematic use of generative AI. Although several professional state bars have issued ethical guidance regarding the use of generative AI and many courts and judges at both the federal and state levels have established standing orders regarding the disclosure of the use of generative AI in pleadings, EOIR has largely lagged behind with guidance for its adjudicatory components. In January 2025 EOIR’s Office of the Chief Administrative Hearing Officer (OCAHO) established general guidelines through adjudication regarding parties’ use of generative AI in its proceedings. See United States v. Wallcon, LLC, 21 OCAHO no. 1630, 9-14 (2025). However, the Office of the Chief Immigration Judge and the Board of Immigration Appeals (BIA) have not—as of yet—established any uniform positions on the subject. EOIR may pursue rulemaking at a future date to provide further guidance on the use of generative AI, and the BIA may also establish guidelines through a published precedential decision. Moreover, the Department of Justice is expected to issue departmentwide guidance on the use of generative AI by the end of 2025. See OMB Memorandum M-25-21, at 12 (requiring each federal agency to “develop a policy that sets the terms for acceptable use of generative AI for their missions and establishes adequate safeguards and oversight mechanisms that allow generative AI to be used in the agency without posing undue risk” within 270 days, or by approximately December 29, 2025). Until further policy directives are issued, however, this Policy Memorandum (PM) provides general guidance for EOIR adjudicators to consider regarding parties’ use of generative AI in immigration proceedings, particularly in cases before the Immigration Courts or the BIA. EOIR has neither a blanket prohibition on the use of generative AI in its proceedings nor a mandatory disclosure requirement regarding its use. Nevertheless, nothing in EOIR’s rules prohibit individual adjudicators or courts from adopting standing orders, see PM 20-09, The Immigration Court Practice Manual and Orders (Feb. 13, 2020), or local operating procedures, see 8 C.F.R. § 1003.40, regulating the use and disclosure of generative AI contained in pleadings. Any such orders or procedures that are adopted remain subject to management approval requirements and will be posted in the appropriate location on EOIR’s website. See PM 20-09 at 3. The BIA, too, may prescribe rules, with the approval of the Director, related to the use of generative AI in pleadings filed with it. See 8 C.F.R. § 1003.1(d)(4). For attorneys who choose to use generative AI tools in the preparation of legal filings in any proceeding before EOIR, that use has the potential to implicate applicable rules of professional conduct, as well as associated ethics rules. See, e.g., American Bar Association, Formal Opinion 512, Generative Artificial Intelligence Tools (Jul. 29, 2024) (ABA Formal Opinion 512), https://www.americanbar.org/content/dam/aba/administrative/professional_responsibility/ethics- opinions/aba-formal-opinion-512.pdf (identifying multiple model rules of professional responsibility implicated by the use of generative AI). Moreover, practitioners appearing before EOIR’s adjudicatory components are expected to act in a professional, ethical manner and in conformance with the applicable rules and standards of professional conduct, including the rules https://www.whitehouse.gov/wp-content/uploads/2025/02/M-25-21-Accelerating-Federal-Use-of-AI-through- Innovation-Governance-and-Public-Trust.pdf. Thus, to the extent that authorized generative AI use is approved in the near future, EOIR adjudicators are strongly cautioned to ensure that any authorized use of generative AI complies with all applicable ethical and professional responsibility obligations. Any improper use of generative AI by an EOIR adjudicator may result in corrective or disciplinary action. This PM is largely consistent with OCAHO’s established guidance. To the extent this PM may conflict with OCAHO’s policies established through adjudication, the OCAHO adjudicatory decisions would control.

of professional conduct of any relevant state bar. See 8 C.F.R. § 1003.101(a) (providing that a practitioner authorized to practice before the Board and the Immigration Courts may be subject to disciplinary sanctions “when such person has engaged in criminal, unethical, or unprofessional conduct, or in frivolous behavior”); 28 C.F.R. § 68.35 (providing that those appearing in proceedings before OCAHO “are expected to act with integrity, and in an ethical manner”). Thus, practitioners submitting hallucinated or erroneous AI-generated content in filings before EOIR likely violate professional conduct obligations, and attorneys who submit such content to an Immigration Court or the BIA may be subject to discipline for “knowingly or with reckless disregard offering false evidence,” 8 C.F.R. § 1003.102(c). Depending on the particular posture and facts of a case and the impact of the filing, the use of such content may also implicate disciplinary rules regarding “[e]ngag[ing] in conduct that constitutes ineffective assistance of counsel,” “[e]ngag[ing] in conduct that is prejudicial to the administration of justice or undermines the integrity of the adjudicative process,” “[f]ail[ing] to provide competent representation to a client,” failing to disclose adverse legal authority, and “[r]epeatedly draft[ing] notices, motions, briefs, or claims that are filed with. . .EOIR that reflect little or no attention to the specific factual or legal issues applicable to a client’s case, but rather rely on boilerplate language indicative of a substantial failure to competently and diligently represent the client.” 8 C.F.R. §§ 1003.102(k)- (o), (s), (u). Consequently, parties and attorneys before each of EOIR’s adjudicatory components should take care to confirm the accuracy of any citations or other research or drafting conducted using generative AI tools. Moreover, in certain circumstances, attorneys may need to consult with their clients regarding the use of generative AI. See ABA Formal Opinion 512 at 8 (discussing situations in which the Model Rules of Professional Responsibility may require the disclosure of the use of generative AI to a client). 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. Accordingly, consistent with EOIR’s core policy values to maintain the integrity of its immigration proceedings, EOIR adjudicators who suspect or discover the inappropriate use of generative AI should report those instances to EOIR’s Attorney Discipline Program and, as appropriate, its Anti-Fraud Program. This PM is not intended to, does not, and may not be relied upon to create, any right or benefit, substantive or procedural, enforceable at law or in equity by any party against the United States, its departments, agencies, or entities, its officers, employees, or agents, or any other person. Nothing herein should be construed as mandating a particular outcome in any specific case. Nothing in this PM limits an adjudicator’s independent judgment and discretion in adjudicating cases or an adjudicator’s authority under applicable law. Please contact your supervisor if you have any questions.