Michaelis v. Wilmington Savings Fund Society: AI-Generated Citations in Pro Se Litigation | Advocate Prakhar

Case at a Glance

Background

Ellen Michaelis, appearing pro se, sued Wilmington Savings Fund Society as trustee for a mortgage acquisition trust, seeking to quiet title to property at 33 Cumberland Road, Fishkill, New York. The property had been foreclosed upon in 2016 following a judgment in a separate action and was subsequently sold. Michaelis claimed a one-fifth inheritance interest under a 1991 will and alleged that a 1992 deed transferring the property was fraudulent. Her filings in support of a vacatur motion included legal citations that the court found unverifiable and bearing hallmarks of AI-generated content submitted without independent verification.

The AI Issue

The court had to determine whether Michaelis’s motion to vacate a prior dismissal order could succeed when her supporting submissions included unverifiable, potentially AI-fabricated legal citations. The broader judicial question was whether pro se litigants who rely on AI tools without verification should be treated differently from attorneys who commit the same error, and what remedy is appropriate when unverified AI citations infiltrate court filings.

What the Court Decided

• The motion to vacate the September 25, 2025 Order was denied in full [no grounds established under CPLR § 5015].
• The Amended Verified Complaint was not accepted for filing, as the action lacked a proper legal and factual basis [failure to state a claim].
• The court observed that filings containing unverifiable citations place an impossible burden on courts and opposing parties [judicial efficiency principle].
• Pro se status does not relieve a party of the duty to verify the accuracy of every legal authority cited in court filings [equal duty of candor].
• The court issued a formal warning that future filings of this nature would attract enhanced scrutiny and potential sanctions.

“It is no more acceptable for a pro se litigant to submit briefs with fake case citations than it is for a lawyer to do so.”

— Dukuray v. Experian Info. Sols., 2024 WL 3812259 (S.D.N.Y. 2024), principle applied by Justice Acker, April 3, 2026

The India Angle

Indian Law Equivalent

In India, submitting fabricated legal citations constitutes a breach of the advocate’s fundamental duty of candor under the Advocates Act, 1961 (Section 35 — professional misconduct). The Code of Civil Procedure, 1908 (Order XI and Section 35B) empowers courts to impose costs on parties who waste judicial time through false or misleading submissions. Additionally, the Information Technology Act, 2000 and the emerging Digital Personal Data Protection Act, 2023 (DPDP Act) frame the accountability landscape when AI tools process legal data without human verification.

Bar Council Rules

Rule 14 of the BCI Standards of Professional Conduct prohibits advocates from making false statements of fact or law before any court. Rule 15 requires that advocates not act in any manner bringing disrepute to the profession or the court. Rule 33 requires advocates to maintain professional conduct at all times. An Indian advocate who submits AI-generated fictitious citations without verification would face proceedings under these rules, potentially leading to suspension or disbarment by the State Bar Council or the Bar Council of India.

Practical Advice for Indian Advocates

• Cross-verify every citation generated by AI against authenticated databases (SCC Online, Manupatra, IndiaKanoon) before filing — no exception, even in urgent matters.
• Treat AI-generated legal text as a first draft only: read the full original judgment yourself before relying on any summary or quoted passage in court submissions.
• Consider adding a verification declaration to AI-assisted filings certifying that all citations have been individually checked, as this practice is fast becoming a judicial expectation globally.

Quick Takeaways

• Pro se litigants bear the same citation accuracy duty as licensed attorneys in court.
• AI-generated filings without verification risk dismissal, warnings, and future sanctions.
• Verify every case citation personally before filing, regardless of which AI tool was used.

Deep Dive: Pro Se Litigants, AI Tools, and the Equal Duty of Candor

The Michaelis case fits within a rapidly expanding body of 2026 jurisprudence in which courts from Florida to New York are confronting a shared crisis: AI tools are producing plausible-sounding but entirely fictitious legal citations, and litigants — both self-represented and legally trained — are submitting them to courts without verification. What makes Michaelis notable is that it is a state court case involving a purely civil property dispute, not a federal high-stakes commercial matter. This demonstrates that the AI hallucination problem has permeated every tier of the judicial system.

The underlying dispute was a quiet title action arising from a 2016 foreclosure on property in Fishkill, New York. Michaelis, acting without counsel, sought to recover an inheritance interest by challenging the validity of a deed from 1992. The procedural history was troubled before AI citations even emerged as an issue: the court had denied a default judgment motion, directed service compliance, and dismissed once before the vacatur motion at issue in this opinion. When Michaelis attempted to reopen the case, she supported her motion with what appeared to be AI-generated legal authorities — a pattern courts in 2025 and 2026 have identified as among the most serious risks of unregulated AI use in litigation.

Compared to landmark cases like Mata v. Avianca, Inc., 678 F. Supp. 3d 443 (S.D.N.Y. 2023), where licensed attorneys were each sanctioned $5,000 for submitting AI hallucinations, Michaelis represents a calibrated judicial response: denial of relief and a formal warning rather than monetary sanctions. This mirrors a trend across 2026 cases where courts are calibrating sanctions based on whether the filer is a lawyer or a self-represented party, whether the misconduct was wilful, and whether a prior warning had been issued. However, courts are unanimous that pro se status is not a shield. As the Eastern District of New York stated in Dukuray v. Experian (2024), cited in related cases, it is “no more acceptable for a pro se litigant to submit briefs with fake case citations than it is for a lawyer to do so.”

For Indian advocates, the lesson is direct and urgent. India’s trial courts and High Courts are increasingly tech-savvy, and a moment of judicial education similar to the 2023-2026 US wave is approaching. Advocates who embed AI verification into their standard workflow now — treating AI as a research assistant, not a final authority — will avoid the professional embarrassment and potential misconduct proceedings that have engulfed numerous US practitioners. The Michaelis case, modest in its direct sanction, is a bellwether for what courts worldwide will expect of all litigants as AI tools become ubiquitous in legal practice.