Amtrust v. Liberty Mutual (NJ 2026): Attorney Fined $1,000 for AI-Hallucinated Cases in Merits Brief | Advocate Prakhar

Case at a Glance

Background

Justin McGinness, employed by Nellies Provisions, was injured on March 28, 2022, when the vehicle he was riding in as a passenger was struck by an at-fault motorist. AmTrust North America, the workers’ compensation carrier, paid $75,339.52 in benefits and recovered only $15,000 from the tortfeasor’s policy limit. AmTrust then sought to recover the remaining $60,339.52 by filing a UIM claim against what it believed was a Liberty Mutual business auto policy covering the employer.

The trial court dismissed AmTrust’s complaint on three grounds: it had named the wrong entity (the proper defendant was Ohio Security Insurance Company, not Ohio Underwriting Managers or Liberty Mutual Insurance Company); it had no public policy right to subrogation on UIM benefits; and a policy exclusion barred recovery for workers’ compensation carriers. AmTrust appealed and filed a merits brief. Defendants’ opposition brief identified four citations in plaintiff’s brief as referring to cases that simply did not exist.

Despite this direct and unambiguous notice, plaintiff’s counsel neither revised the filed brief, filed a corrective disclosure, nor informed the court of the error before the Appellate Division decided the case. Both opposing counsel and the court independently spent time researching non-existent authorities.

The AI Issue

The Appellate Division characterized the situation as the “presumed misuse of artificial intelligence,” noting that the pattern of four completely fabricated citations — plausible in form but non-existent in substance — was consistent with generative AI hallucination. The court identified a two-part legal problem: first, whether filing fake citations violates the duty of candor under RPC 3.3; and second, whether failing to correct those citations after receiving express notice from opposing counsel constitutes an independent and aggravating violation. The court answered yes to both.

What the Court Decided

• Filing four non-existent case citations in an appellate merits brief violates New Jersey RPC 3.3, which prohibits false statements of law to a tribunal.
• Rule 2:9-9 of the New Jersey Court Rules imposes independent obligations of accuracy in appellate filings; fabricated citations violate this rule.
• The failure to act after being specifically notified by opposing counsel of the fake citations was an independent and aggravating violation — counsel had the means and opportunity to correct the record and chose not to.
• Both opposing counsel and the court wasted research time on non-existent authorities, establishing the prejudice necessary for a sanctions award.
• A $1,000 sanction was imposed on plaintiff’s counsel personally, with express ethical violation findings placed on the record.

“Counsel’s disregard for his obligations toward his adversary and this court is a violation of RPC 3.3 and Rule 2:9-9, particularly in light of counsel’s failure to react when alerted to the error.”

— New Jersey Appellate Division, Amtrust v. Liberty Mutual, March 27, 2026

The India Angle

Indian Law Equivalent

Indian courts treat persistent reliance on fabricated legal authority as contempt. Under the Supreme Court Rules, 2013, Order IV, counsel has a continuing duty to correct any misstatement made to the court once the error comes to their attention. The Supreme Court in Mahipal Singh Rana v. State of UP (2016) 8 SCC 335 held that the advocate’s duty of candour is absolute and continuing. An advocate who knows their brief contains false citations and does nothing violates this duty at every subsequent hearing.

Bar Council Rules

BCI Rules, Chapter II, Rule 9 prohibits an advocate from doing anything that tends to mislead the court. Rule 22 requires that facts stated in pleadings and representations be true to the best of the advocate’s knowledge. The Amtrust scenario — where counsel persists in relying on non-existent authority after being specifically alerted — violates both rules and would warrant disciplinary proceedings under Section 35 of the Advocates Act, 1961.

Practical Advice for Indian Advocates

• When opposing counsel identifies a citation error that may be AI-generated, investigate immediately — treat the notice as a formal trigger of your continuing duty to correct misrepresentations.
• If the citation is confirmed fabricated, file a correction memo before the next hearing date, which substantially reduces disciplinary exposure.
• Build a pre-filing checklist for AI-assisted briefs that requires a colleague cross-verification step for all cited authorities before signing and filing.

Quick Takeaways

• Failing to correct known AI hallucinations after notice is independently sanctionable conduct.
• $1,000 fine signals NJ courts treat AI-citation ethics seriously even in commercial matters.
• The duty to correct a known misrepresentation is continuing — inaction is never a neutral choice.

Deep Dive: The Post-Notice Failure Standard in AI Sanction Cases

Amtrust v. Liberty Mutual contributes a critical dimension to the AI hallucination sanctions landscape: what happens when an attorney learns about fake citations through the adversarial process rather than through self-discovery? Prior landmark cases like Mata v. Avianca focused primarily on the original filing error. Amtrust establishes that a second, independent violation arises when counsel receives express notice and fails to act.

The court’s reasoning is grounded in the continuing nature of RPC 3.3. The rule does not merely prohibit making false statements — it imposes an ongoing obligation to correct material misrepresentations once the attorney is aware of them. The moment opposing counsel’s brief identified the fake citations, plaintiff’s attorney was already under a duty to correct the record. His silence was not a neutral act; it was a fresh violation compounding the original one.

This has significant practical implications for AI-assisted drafting workflows. Even in a scenario where the original filing might theoretically be characterized as an honest mistake, any window for mitigation closes the instant opposing counsel raises the specific issue. Attorneys who receive a challenge to AI-generated citations must treat that challenge as presumptively credible and investigate immediately rather than waiting to see whether the court notices independently.

For Indian advocates appearing before High Courts and the Supreme Court, the Amtrust model provides a clear behavioral framework. The moment an error is pointed out, the advocate faces a choice between corrective transparency and doubled liability. The New Jersey court’s $1,000 sanction — while modest in amount — was accompanied by express ethical violation findings that become part of the permanent professional record, a consequence that far outweighs the monetary penalty itself.