Rubio v. Mullin: AI Hallucinations in Immigration Court | Advocate Prakhar

Case at a Glance

Background

Yasmani Gurri Rubio, a Cuban national proceeding without a lawyer, sued the Secretary of Homeland Security after USCIS failed to decide his naturalization application within 120 days of his April 2025 interview. Under 8 U.S.C. § 1447(b), such delay gives federal courts jurisdiction to determine the matter or remand it. Rubio had already filed nine motions by the time the court issued its May 2026 omnibus opinion — a volume suggesting that AI-assisted drafting may have lowered the practical barrier to litigation but not the accuracy requirement.

The AI Issue

Rubio’s complaint quoted Tutun v. United States, 270 U.S. 568, 578 (1926) as saying: “When the applicant shows his eligibility, the court has the duty to grant the application.” The judge verified that no such language exists anywhere in Tutun. Worse, the actual holding of that case is precisely the opposite — naturalization is a privilege, not a right. In a later filing, Rubio also attributed to the District of Massachusetts a quotation from Taalebinezhaad v. Chertoff that similarly does not appear in that decision. The pattern strongly suggested AI-generated text presented as real judicial authority.

What the Court Decided

• Non-existent quotations “cause deep concern” and “significantly weaken the Court’s trust” — the court explicitly named the fabrication problem. [candor to the tribunal]
• Rule 60(b)(1) relief denied because denial of a default motion is not a “final judgment.” [finality doctrine]
• Case reopened under Rule 60(b)(6) because voluntary dismissal was without prejudice. [equitable relief]
• Matter remanded to USCIS under 8 U.S.C. § 1447(b) with a 90-day adjudication deadline. [primary jurisdiction]
• Court warned that future fabricated citations in any court “may be sanctionable.” [prospective deterrence]

“Mr. Gurri Rubio’s use of nonexistent quotes causes deep concern to the Court, wastes the Court’s and opposing counsel’s time and resources, and significantly weakens the Court’s trust in his papers.”

— Judge Adam B. Abelson, D. Maryland, May 1, 2026

The India Angle

Indian Law Equivalent

In India, knowingly false statements of law before a court attract criminal contempt under Section 2(d) of the Contempt of Courts Act 1971. Filing fabricated citations also violates Section 191 IPC (false evidence) and could trigger proceedings under Order XI CPC. Indian courts dealing with citizenship delays under the Citizenship Act 1955 and the Foreigners Act 1946 are equally strict — writ petitions must cite authentic Supreme Court or High Court authority verifiable on the official SCI database.

Bar Council Rules

Rule 32 of the Bar Council of India Rules (Advocates Act 1961) prohibits any act involving fraud or deceit, and Rule 33 requires fair representation. Submitting AI-fabricated legal citations violates both rules and exposes advocates to disciplinary action before the State Bar Council, suspension, or even removal from the roll under Section 35 of the Advocates Act.

Practical Advice for Indian Advocates

• Verify every citation through SCC Online, Manupatra, or the Supreme Court of India website — never rely on AI output as the primary source for case quotations.
• When advising clients on OCI card rejections or citizenship disputes, ensure all MHA circulars cited are pulled directly from the official Gazette of India, not AI summaries.
• Instruct self-represented clients who draft their own petitions to submit drafts for citation verification before filing — the Rubio case shows pro se status offers no immunity from sanction.

Quick Takeaways

• Pro se litigants face sanctions for AI hallucinations just like licensed attorneys.
• Misrepresenting a case to mean the opposite of its actual holding is especially damaging to credibility.
• Courts will remand rather than reward litigants who combine procedural over-filing with fabricated citations.

Deep Dive: When Self-Represented Litigants Use AI — The Growing Crisis in Immigration Courts

The Rubio v. Mullin decision occupies a distinctive corner of the AI hallucination jurisprudence landscape: it involves a self-represented foreign national navigating naturalization proceedings rather than a licensed attorney. Most earlier AI-sanction cases targeted trained lawyers, creating a perception that the hallucination crisis was primarily a problem of professional negligence. Rubio confirms the problem is systemic — anyone who generates court filings using unchecked AI output risks judicial sanction.

What makes this case qualitatively different is not just that Rubio cited a non-existent quotation, but that the actual precedent he cited — Tutun v. United States from 1926 — says the precise opposite of what he claimed. Generative AI models, trained on vast legal corpora, sometimes produce plausible-sounding quotes that blend the correct case name and citation with invented language. The result is a document that passes a superficial citation check but fails the moment a judge actually reads the source. This is not a new problem in AI research, but Rubio demonstrates its real-world consequences in the highest-stakes setting possible: a person’s eligibility to become a citizen.

The volume of Rubio’s filings — nine motions in roughly eight months — also illustrates a concerning pattern. AI tools lower the practical cost of drafting legal arguments, which can encourage self-represented litigants to file more motions than the case warrants. This creates compounding problems: courts spend disproportionate resources evaluating borderline or frivolous arguments, and litigants whose AI-generated briefs contain false authority undermine their own substantively meritorious positions. In Rubio’s case, his underlying claim — that USCIS delayed his application beyond the statutory 120-day window — had legal merit. It was his fabricated citations that eroded judicial sympathy and trust.

For Indian immigration practitioners, the parallel is instructive. India processes hundreds of thousands of visa, OCI, and citizenship applications annually, and High Court writ petitions challenging Ministry of Home Affairs delays are increasingly common. If advocates or their clients use AI tools to draft such petitions without verifying every citation through authoritative databases, they risk the same outcome as Rubio: a sympathetic underlying claim undermined by fabricated authority. The safer approach is to treat AI output as a first draft requiring full legal verification — a research assistant, not a research authority.