⚡ Case Digest
FECTEAU v. SAFETY NATIONAL CASUALTY CORP. — U.S. District Court, S.D.N.Y., March 25, 2026
Pro se plaintiff Matthew Fecteau invented non-existent statutory language from the SCRA and fabricated quotes from real US Supreme Court cases in insurance coverage filings, prompting Judge Karas to impose a filing injunction requiring prior leave of court for all future submissions against the defendant.
Why it matters: Courts treat fabricated statutory text and false quotes from real cases as the most serious form of AI hallucination misconduct, justifying a permanent-style filing injunction even against pro se litigants.
Category: AI Hallucination & Sanctions | Jurisdiction: USA (New York Federal) | Read time: 6 min
Case at a Glance
| Full Citation | Matthew J. Fecteau v. Safety National Casualty Corp. et al., No. 25-CV-3821 (KMK), (S.D.N.Y. Mar. 25, 2026) |
| Court | U.S. District Court, Southern District of New York |
| Date | March 25, 2026 |
| Category | Fabricated Quotes / AI Hallucination / Filing Injunction |
| Jurisdiction | United States (New York Federal) |
| AI Tool Used | Inferred AI (fabricated statutory and case language without acknowledgement) |
| Outcome/Sanction | All claims dismissed without prejudice; filing injunction against further submissions without leave of court |
Background
Matthew Fecteau, a pro se plaintiff and military servicemember, sued Safety National Casualty Corporation and Sedgwick Claims Management Services for bad faith insurance claims handling and violations of New York General Business Law § 349, arising from his civil rights claim in a related action. Fecteau filed an extraordinarily high volume of motions — including multiple remand motions, motions invoking the Servicemembers Civil Relief Act (SCRA), and various supplemental submissions — many of which contained language that was quoted in quotation marks but did not appear in the cited sources.
The AI Issue
The court had to determine whether Fecteau’s filings, which contained fabricated statutory language purportedly from the SCRA and fabricated quotes from real US Supreme Court decisions including Conroy v. Aniskoff, constituted sanctionable misconduct under Rule 11 and/or the court’s inherent authority. The fabrication of text placed in quotation marks and attributed to real sources represents the most deceptive form of AI hallucination behaviour — worse than citing a non-existent case, because it exploits real authority while attributing false content to it.
What the Court Decided
- Defendants’ Motions to Dismiss were granted in full; all claims dismissed without prejudice [case dismissed].
- Safety National’s Motion for Sanctions was granted in part: Fecteau invented statutory language attributed to the SCRA and fabricated quotes from Conroy v. Aniskoff and Gordon [fabrication of primary sources confirmed].
- A filing injunction was imposed temporarily enjoining Fecteau from pursuing claims against Safety National without prior leave of court [filing injunction under Rule 11(c)(1) and (4) and inherent authority].
- The court held that “placing invented language in quotation marks and representing that it appears in a specific source” is fabrication, not ordinary legal drafting [definition of fabrication established].
- Pro se status does not lower the standard for candour; fabricated quotes from real cases are as unacceptable as entirely fictitious citations [equal standard confirmed].
“The Court cannot imagine how invented quotes could be ‘grounded in the statutory text,’ and notes that, under both the Federal Rules of Civil Procedure and the basic conventions of written English, it is not ‘ordinary drafting’ (legal or otherwise) to invent language, place it in quotation marks, and represent that it appears in a specific source. It is, in fact, ‘fabrication.'”
— Judge Kenneth M. Karas, S.D.N.Y., March 25, 2026
The India Angle
Indian Law Equivalent
In India, fabricating quotes from real judicial decisions and attributing them to existing authorities would constitute criminal contempt under the Contempt of Courts Act, 1971 (Section 2(c)) and professional misconduct under the Advocates Act, 1961. Section 193 of the Indian Penal Code (now Section 229 BNS 2023) criminalises giving false evidence, including making false statements in legal proceedings. Fabricating statutory text and attributing it to a real statute could also engage Section 464 IPC (making a false document) and its BNS equivalent.
Bar Council Rules
BCI Rule 14 (no false statements to court), Rule 15 (no disrepute to profession), Rule 22 (maintain dignity of proceedings), and Rule 33 (professional conduct at all times) are all directly engaged by fabrication of quoted statutory and case text. Indian Bar Councils have the power to suspend or debar advocates for such conduct under Section 35 of the Advocates Act; the Supreme Court also has independent powers under the Contempt of Courts Act to address false statements in proceedings before it.
Practical Advice for Indian Advocates
- Never place text in quotation marks in a court submission unless you have personally read the exact words in the original source document — AI tools regularly generate plausible-sounding but false quotes from real cases.
- When using AI to draft submissions, strip out all quotation marks and treat all quoted material as paraphrase until you have personally verified the exact text in the original judgment or statute.
- Fabricating statutory language in India — even unintentionally through AI reliance — can result in criminal contempt proceedings and disbarment; the professional and personal stakes are vastly higher than a monetary fine.
Quick Takeaways
- Fabricating quotes from real cases is the most serious form of AI hallucination misconduct.
- Filing injunctions are available against pro se litigants who engage in repeated fabrication.
- Never use quotation marks in court filings without personally verifying the exact text.
Deep Dive: The Hierarchy of AI Misconduct — Fabricated Quotes vs. Fictitious Citations
Within the rapidly developing AI sanctions jurisprudence of 2025-2026, courts are implicitly establishing a hierarchy of misconduct severity. At the least severe end sits the unintentional citation of a non-existent case — a genuine AI hallucination where the filer relied on an AI tool’s output without verification. Moving up the hierarchy is citing real cases for propositions they do not support, followed by quoting real cases with text that does not appear in them. At the most severe end — where Fecteau sits — is the fabrication of statutory text attributed to real, named legislation, combined with the attribution of false quotes to real, named Supreme Court decisions.
Judge Karas’s articulation in Fecteau is analytically important: “placing invented language in quotation marks and representing that it appears in a specific source” is not a grey area or a matter of interpretation. It is fabrication. This formulation is likely to be cited extensively by courts across the US as they develop their own frameworks for distinguishing degrees of AI misconduct. The court’s refusal to accept Fecteau’s characterisation of his fabricated SCRA quotes as “ordinary legal drafting” or “paraphrasing” forecloses a defence that future litigants might attempt to invoke.
The filing injunction imposed in Fecteau is also noteworthy because it applies not just to future AI-assisted filings but to all further submissions against Safety National, whether in federal or state court. This extra-territorial reach — explicitly authorised by the Second Circuit under the Anti-Injunction Act exceptions — represents a significant escalation of the sanction beyond the typical monetary fine. It treats Fecteau’s fabrication as a systemic risk to judicial integrity rather than a one-off mistake deserving correction.
For Indian advocates, the Fecteau case contains a specific and immediately actionable warning. AI tools — particularly large language models trained on legal text — frequently generate plausible-sounding quotes attributed to real cases but not found in those cases. The AI tool is not lying in a legal sense; it is producing statistically probable text in the style of the cited source. But when an advocate or litigant places that text in quotation marks and files it with a court, they have crossed the line from inadvertent error to fabrication, regardless of whether they subjectively intended to deceive. Indian courts have not yet encountered this specific fact pattern in reported decisions, but the scenario is entirely predictable as AI use in Indian legal practice grows.