Is This AI Rule Necessary? An Analysis of Assembly Bill 8546

Is This AI Rule Necessary? An Analysis of Assembly Bill 8546

7.2.2026

By Alexander Paykin

In June 2023, in a Pearl Street courtroom in Manhattan, Judge P. Kevin Castel sat down to write what would become one of the most quoted federal opinions of the decade, Mata v. Avianca, Inc.[1] The facts have entered the legal malpractice canon. A New York personal injury attorney typed his client’s case into ChatGPT, asked the chatbot for supporting authority, and got six elegantly captioned, internally cited, and entirely fabricated cases. He filed them. When opposing counsel and the court could not find them, he filed an affirmation attaching the decisions, which were also fabricated. Judge Castel imposed a $5,000 sanction on the lawyers and the firm and ordered letters be sent to each of the real federal judges falsely identified as the authors of the bogus opinions.

Seven months later, on Jan. 30, 2024, the Second Circuit referred attorney Jae S. Lee to its grievance panel for citing in a reply brief a single non-existent case generated by ChatGPT.[2] One year after that, in February 2025, Magistrate Judge Kelly Rankin in Wyoming sanctioned three attorneys from Morgan & Morgan, “America’s largest injury law firm,” for citing eight fabricated cases produced by their firm’s in-house artificial intelligence tool, MX2.law.[3] Then, in January 2026, the Appellate Division, Third Department, in Deutsche Bank National Trust Co. v. LeTennier, imposed monetary sanctions on counsel and a litigant for an opening brief and successive filings totaling 23 fabricated cases. Commentators called it “the first appellate-level case in New York” to do so.[4]

This is the public record on which the Legislature seeks to act. As of this writing, the Damien Charlotin AI Hallucination Cases Database has identified more than 1,394 cases worldwide in which generative AI produced fabricated content in legal filings.[5] The number grows by the week. Against this backdrop, the New York Legislature has introduced Assembly Bill A.8546 (Lavine), with Senate companions S.9794 (Sepúlveda) and S.2698 (Hoylman-Sigal), proposing a mandatory AI-disclosure affidavit requirement for court filings. [6] This article examines the bill’s provisions, its relationship to the existing enforcement framework, the definitional challenges it presents, comparative approaches in other jurisdictions, the recent OCA Interim Policy, and the practical implications for New York practitioners.

What A.8546 Would Do

Strip A.8546 to its bones and it does three things. First, it adds a new CPLR Rule 2107 requiring that “any paper or file served that was drafted with the assistance of generative artificial intelligence must attach to the filing a separate affidavit disclosing such use and certifying that a human being has reviewed the source material and verified that the artificially generated content is accurate.” Second, it amends CPLR 5528(a) to extend the same requirement to appellate briefs. Third, it takes effect 90 days after enactment.

The central issue is the definition. Subsection (b) defines “generative artificial intelligence” as “the use of machine learning technology, software, automation, and algorithms” performing tasks, making rules or predictions based on data, including any system performing tasks “under varying and unpredictable circumstances without significant human oversight, or that can learn from experience”; any system addressing tasks requiring “human-like perception, cognition, planning, learning, communication, or physical action”; systems “designed to think or act like a human, including cognitive architectures and neural networks”; “any set of techniques, including machine learning, designed to approximate a cognitive task”; and systems “designed to act rationally, including an intelligent software agent or embodied robot.”

That is not the definition of generative artificial intelligence. That is the definition of artificial intelligence, and not even narrowly. As Carolyn Elefant pointed out in her now famous public comment to the 5th Circuit’s parallel proposal in 2024, the definition cannot tell a lawyer “when she ‘uses’ generative AI. Is it when she considers a computer’s proposed language during a search and rejects it? Or incorporates it in a draft but then writes over it in later edits, keeping just a few words?” [7]

What New York Already Does

Before considering whether A.8546 is necessary, ask the basic question: What would happen to a lawyer who today filed a brief like the Mata brief in a New York Supreme Court?

Six things, all of them already on the books.

First, the lawyer’s signature on the paper would constitute a certification under 22 N.Y.C.R.R. Section 130-1.1(a) that, “to the best of that person’s knowledge, information and belief, formed after an inquiry reasonable under the circumstances, the presentation of the paper or the contentions therein are not frivolous. [8] Filing a hallucinated case fails that certification on its face.

Second, frivolous conduct under 22 N.Y.C.R.R. Section 130-1.1(c), which includes “asserting material factual statements that are false,” exposes the lawyer to financial sanctions. [9]

Third, Judiciary Law Section 487, a statute traceable to the First Statute of Westminster of 1275, [10] provides treble damages and misdemeanor liability for any attorney who “is guilty of any deceit or collusion … with intent to deceive the court or any party. [11] Citing a fabricated case to a court with knowledge of (or reckless disregard for) its fictional character is the textbook fact pattern.

Fourth, Rule 3.3(a)(1) of the New York Rules of Professional Conduct prohibits a lawyer from “knowingly” making “a false statement of fact or law to a tribunal.” Rule 3.3(a)(3) requires the lawyer to take reasonable remedial measures upon discovery of a false statement. Rule 1.1 (Competence), illuminated by a New York City Bar Association opinion, [12] the NYSBA Task Force Report on Artificial Intelligence (April 6, 2024), [13]and ABA Formal Opinion 512 (July 29, 2024), already requires lawyers to understand the technology they use, including its propensity to “hallucinate.” Rule 5.3 makes the lawyer responsible for supervising AI tools as nonlawyer assistants. Rule 8.4(c) prohibits dishonesty.

Fifth, in any tort action, CPLR 8303-a permits the imposition of attorney fees up to $10,000 for frivolous conduct. And sixth, under CPLR 3126, courts may strike pleadings for willful failure to comply with disclosure obligations.

All of that machinery actually works. In LeTennier, the Third Department reached for 22 N.Y.C.R.R. Section 130-1.1 and obtained sanctions without any new statute. In Mata, Judge Castel reached for Rule 11 and obtained $5,000. In Park, the Second Circuit referred the offending lawyer to its grievance panel under Local Rule 46.2. In People v. Crabill, the Colorado Office of the Presiding Disciplinary Judge reached for the existing Rules of Professional Conduct and suspended a lawyer for one year and a day. [14] In United States v. Hayes, a federal magistrate judge reached for the inherent authority of the federal courts and imposed a $1,500 sanction on an assistant federal defender. [15]

In each of these cases, courts applied existing rules without new AI-specific legislation. The enforcement record raises the question whether A.8546 addresses a gap that current rules leave open.

THE ENFORCEMENT VOID A.8546 contains: • No civil penalty. • No private right of action. • No designated enforcer. • No safe harbor or cure period. • No scienter requirement. • No exemption for de minimis assistance. • No protection for confidentiality or work product. • No interaction provision with 22 N.Y.C.R.R. Part 130, CPLR 8303-a, or Judiciary Law Section 487. • No transitional rule. • No fiscal note (Sponsor’s Memo: “None”). The closest thing to a penalty is perjury liability under Penal Law Sections 210.10 and 210.15 for a knowingly false certification. That liability is nominal. An attorney who simply omits the affidavit faces no perjury exposure. The bill thus creates a disclosure obligation without a corresponding enforcement mechanism for non-compliance.

The Definition Problem

I have a colleague who is bilingual. Sometimes he uses Google Translate to spot-check a Russian term in a translation he has commissioned. Google Translate is a deep-learning neural machine translation system. It learns from data. It performs the cognitive task of translation. It operates without significant human oversight. On the face of A.8546’s subsection (b), it is “generative artificial intelligence.”

Microsoft Word’s grammar suggestions and its Editor and Copilot features are machine-learning systems performing the cognitive task of editing. They are generative AI under the bill. So is Outlook Smart Compose. So is Apple Intelligence’s Writing Tools. So is Grammarly. So is iOS predictive text.

Adobe Acrobat OCR, used to convert a faxed exhibit to a searchable PDF, is a machine-learning perception system. It is generative AI under the bill.

Westlaw KeyCite, Lexis BCite, and Bloomberg BCite, all of which perform machine-learning analysis of citation networks to flag bad law, are within the definition. Westlaw Precision, Westlaw CoCounsel, and Lexis+ AI explicitly are. So are Casetext, Harvey, Hebbia, Paxton, Spellbook, and Bloomberg Law’s AI Assistant.

Every modern e-discovery technology-assisted review workflow is squarely within subsections (b)(1) and (b)(4): “any artificial system that … can learn from experience and improve performance when exposed to data sets” and “any set of techniques, including machine learning, designed to approximate a cognitive task.” Predictive coding has been in commercial use in this state since at least Da Silva Moore v. Publicis Groupe. [16] It is now standard practice.

Even the calendar in your phone, which uses machine learning to suggest meeting times, meets the definition.

The New York City Bar Association saw this clearly. Footnote 21 of Formal Opinion 2024-5 says it plainly: “Some Generative AI is routinely used. For example, Microsoft Word employs Generative AI in its auto-complete and grammar check functions. Westlaw, Lexis, and search engines also employ Generative AI. We do not mean to suggest that an attorney needs to disclose such uses of Generative AI.” [17]

A.8546’s drafters did not draw that line. The bill’s definition draws no line at all.

The vagueness problem is not academic. Under Grayned v. City of Rockford, and FCC v. Fox Television Stations, Inc., [18] a regulation that fails to “give a person of ordinary intelligence fair notice of what is forbidden” and “invites arbitrary and discriminatory enforcement” raises constitutional concerns. A definition that captures the spell-check on the brief, optical character recognition on the exhibit, the calendar on the lawyer’s phone, and the autocomplete on the email transmitting service is a definition that delegates to every individual judge the unconstrained discretion to decide what counts.

What Other Jurisdictions Have Done, and Not Done

The federal experience is instructive. Beginning in May 2023, Judge Brantley Starr of the Northern District of Texas issued a much discussed standing order requiring counsel to certify either non-use of generative AI or human verification of any generative AI text. [19] By February 2024, Law360 reported that approximately 16% of federal district and magistrate judges had issued some form of AI standing order. Most have since rolled them back, narrowed them, or quietly let them lapse.

The most consequential federal action came on June 12, 2024. After notice and comment on a proposed amendment to 5th Circuit Rule 32.3 and Form 6, which would have required certification that “no generative artificial intelligence program was used in drafting the document presented for filing, or to the extent such a program was used, all generated text, including all citations and legal analysis, has been reviewed for accuracy and approved by a human,” the U.S. Court of Appeals for the 5th Circuit formally declined to adopt the rule. The court stated: “The court … has decided not to adopt a special rule regarding the use of artificial intelligence in drafting briefs at this time. Parties and counsel are reminded of their duties regarding their filings before the court … ‘I used AI’ will not be an excuse for an otherwise sanctionable offense.” [20]

The public comments to the 5th Circuit, from a wide cross-section of the bar, are worth reading. Layne Kruse and Warren Huang of Norton Rose Fulbright described the proposed rule as “unnecessary.” Carolyn Elefant warned of work-product invasion. David Coale highlighted definitional indeterminacy. Brian King wrote that “AI-focused court rules” risk being “a passing fad.” [21]

The international comparators are also instructive. The Federal Court of Canada, in its Notice to the Parties and the Profession on the Use of Artificial Intelligence in Court Proceedings (Dec. 20, 2023, revised May 2024), requires a declaration only when generative AI was used to “create or generate” content. It explicitly carves out tools that “merely suggest changes, provide recommendations, or critique content already created by a human.”[22] That narrower approach illustrates one alternative definitional framework.

In R (Ayinde) v. London Borough of Haringey and Al-Haroun v. Qatar National Bank QPSC, Dame Victoria Sharp P. and Justice Johnson, in the Divisional Court, addressed two cases involving fabricated citations. They held that existing duties to the court and the regulators (the Bar Standards Board and the Solicitors Regulation Authority) provided sufficient discipline. The court declined to mandate disclosure as a categorical matter, instead emphasizing that “the administration of justice depends upon the court being able to rely without question on the integrity of those who appear before it.”[23]

In Canada, Zhang v. Chen applied existing rules to order a lawyer personally to bear opposing counsel’s costs for fabricated ChatGPT cases. [24] In Australia, Valu v. Minister for Immigration did the same. The EU AI Act’s Article 50 transparency obligations and Annex III’s classification of AI in administration of justice as high-risk address developers and deployers, not litigant disclosure, and so are not the right comparator for a procedural disclosure rule.

The pattern across jurisdictions is notable: Courts and regulators that have considered AI-specific disclosure mandates have generally concluded that existing rules of court, professional responsibility and judicial inherent authority address the identified harms. Several early disclosure mandates have since been narrowed or withdrawn.

What New York Courts Have Done

A significant development occurred after the bill’s introduction. After 18 months of deliberation, the Office of Court Administration Advisory Committee on Artificial Intelligence and the Courts, established by Chief Judge Rowan Wilson and Chief Administrative Judge Joseph Zayas, addressed substantially the same subject matter as A.8546 through administrative action.

On Oct. 10, 2025, the Unified Court System issued its Interim Policy on the Use of Artificial Intelligence governing judges and non-judicial staff. The policy framework expressly states that attorneys “should not be required, upon submitting papers, to disclose to the court that they have used AI in the preparation of such papers,” because “those duties and responsibilities already apply to all submissions, regardless of whether AI tools were used.” [25] An optional model part rule may be adopted by individual judges at their discretion. It restates 22 N.Y.C.R.R. Section 130-1.1a’s certification requirement and confirms that an attorney “is required to carefully review the paper and independently ensure that it contains no fabricated or fictitious cases, statutes, or other material.”

If A.8546, S.9794, and S.2698 are enacted, they would supersede the Office of Court Administration’s framework by statute, with a definition broader than anything it, the New York City Bar, the city bar’s AI task force, or the NYSBA task force has proposed. The bills were introduced without consultation with the Office of Court Administration, without a fiscal note, without a hearing record, and with an effective date inconsistent with the operative date of the Unified Court System’s policy. Practitioners should be aware that the statutory and administrative frameworks, as drafted, are not aligned.

Update: Part 161 Effective June 1, 2026

What was an optional model part rule has now become a formal, statewide rule of court. Effective June 1, 2026, the Unified Court System adopted Part 161 of the Rules of the Chief Administrative Judge (22 N.Y.C.R.R. Part 161), regulating the use of artificial intelligence by attorneys in the New York State courts. Part 161 carries forward the approach of the October 2025 Interim Policy: it permits the use of AI tools in preparing submissions and does not mandate disclosure of that use. Instead, it provides that attorneys using such tools are “required to carefully review the paper and independently ensure” that their AI-assisted submissions contain no “fabricated or fictitious cases, statutes, or other material.” The rule is expressly grounded in 22 N.Y.C.R.R. Section 130-1.1 and Rule 3.3 of the New York Rules of Professional Conduct, which already prohibit a false material factual statement to the court and the advancement of a frivolous legal argument. Although Part 161 applies statewide, individual judges retain discretion to implement their own AI-related part rules, adopt the model rule set forth in Part 161, or impose no additional requirements.

Part 161 sharpens, rather than resolves, the divergence between the administrative and the proposed statutory frameworks. The two differ in at least four respects. First, on disclosure: Part 161 affirmatively declines to require attorneys to disclose AI use, whereas A.8546 would compel a separate disclosure affidavit attached to every AI-assisted filing. Second, on definitional scope: Part 161 does not import A.8546’s sweeping definition of “generative artificial intelligence” and instead ties an attorney’s obligation to the substance of the submission (no fabricated or fictitious authority), avoiding the over-inclusion problem that captures spellcheck, OCR, translation, and predictive coding. Third, on legal mechanism: Part 161 is a court rule promulgated under the Chief Administrative Judge’s existing rulemaking authority and rests on the duties already codified in Part 130 and Rule 3.3, while A.8546 would create new statutory provisions (CPLR 2107 and an amended CPLR 5528(a)). Fourth, on flexibility: as an administrative rule, Part 161 can be amended as the technology evolves and expressly preserves individual judges’ discretion over their own part rules, whereas a statute would fix the requirement in the CPLR and bind all courts uniformly.

The practical upshot is that, as of June 1, 2026, New York already has a statewide rule directly addressing AI-assisted filings, and it does so without the disclosure affidavit, the broad definition, or the statutory rigidity of A.8546. If the Legislature enacts A.8546, S.9794, and S.2698, the resulting statutory affidavit requirement would sit atop Part 161 rather than integrate with it, leaving practitioners to reconcile a no-disclosure court rule with a mandatory-disclosure statute. The two frameworks, as drafted, point in different directions.

Practical Considerations for New York Practitioners

Whatever the legislative outcome, practitioners navigating AI-assisted legal work in New York face several practical issues arising from the interplay between A.8546, the Office of Court Administration Interim Policy and existing rules.

1. Understanding the Definitional Scope

Under A.8546 as drafted, the definition of “generative artificial intelligence” is broad enough to encompass spelling and grammar tools, citation-formatting and citation-verification tools, optical character recognition, translation software, predictive coding and technology assisted review, calendar and scheduling software, and search engines. Practitioners using any of these tools in connection with a filing would need to assess whether disclosure is triggered. By contrast, the Canadian federal court model limits the obligation to AI used to “create or generate” substantive content, and the New York City Bar’s Formal Opinion 2024-5 expressly excludes routine tools. Lawyers should track which definitional framework applies in each forum.

2. The Existing Part 130 Certification Obligation

Regardless of whether A.8546 is enacted, practitioners are already certifying under 22 N.Y.C.R.R. Section 130-1.1a that their filings are not frivolous and that material factual statements are not false. This certification attaches to every signed filing. Lawyers using AI tools to assist in drafting should ensure their verification workflows satisfy this existing reasonable inquiry standard, as the Mata, Park, and LeTennier sanctions all rested on these pre-existing obligations.

3. The CPLR 2106 Affirmation Framework

The 2024 amendment to CPLR 2106, as further refined by the November 2025 amendment, created a unified affirmation regime under penalty of perjury available to all signers. Practitioners should note that A.8546 requires a “separate affidavit” rather than incorporating disclosure into the existing CPLR 2106 affirmation. If enacted, this would add a standalone procedural requirement distinct from the affirmation already accompanying filings.

4. Verification and Correction Practices

Under Fed. R. Civ. P. 11(c)(2), diligent verification followed by prompt withdrawal or correction defeats sanction. A.8546, by contrast, contains no safe-harbor provision. Practitioners should develop and document internal verification workflows for AI-assisted drafting, including independent citation checking and factual verification, both to satisfy current Part 130 obligations and to establish a record of reasonable inquiry in the event of an inadvertent error.

5. The Pro Se Litigant Dimension

The Charlotin database indicates that pro se litigants account for the majority of AI-tainted filings in U.S. courts. A.8546 applies to “any paper or file served,” which on its face extends to pro se filings, though the bill’s certification mechanism is drafted around attorney obligations. Practitioners appearing opposite pro se parties should be aware that the enforcement landscape for AI-generated filings may differ depending on whether the filer is represented.

6. Work Product and Confidentiality Concerns

New York Rules of Professional Conduct 1.6 and Hickman v. Taylor (1947),[26] protect attorney mental impressions and client confidences. If A.8546’s disclosure requirement is interpreted to require identification of which portions of a filing were AI-assisted and what prompts were used, practitioners face potential tension with work-product doctrine and the duty of confidentiality. The bill contains no express provision addressing this intersection. Lawyers should consider how to comply with any disclosure obligation without revealing protected mental impressions or confidential information.

7. The Pace of Technological Change

AI capabilities evolve significantly faster than statutory frameworks. Practitioners should monitor whether tools that fall outside the bill’s definition today may fall within it tomorrow, and whether tools currently captured may become so ubiquitous as to make categorical disclosure impractical. The Office of Court Administration Interim Policy, as an administrative framework, is more readily updated than a statute.

8. Available Guidance

Practitioners seeking to develop responsible AI-use practices have substantial guidance available. The NYSBA April 2024 Task Force Report, the New York City Bar’s Formal Opinion 2024-5, the Office of Court Administration Advisory Committee’s 2025 report, ABA Formal Opinion 512, and the Unified Court System Interim Policy each address competence, supervision and verification duties. These authorities collectively provide a framework for responsible AI use that does not depend on whether A.8546 is enacted.

Conclusion

A.8546 addresses a real phenomenon: AI-generated fabrications in court filings are occurring with increasing frequency, as the Charlotin database and the decisions in Mata, Park, Wadsworth, and LeTennier confirm. The analytical question is whether the bill’s mechanism – a mandatory separate affidavit, triggered by a definition that encompasses virtually all machine-learning tools – adds enforcement capacity beyond the framework already in place through 22 N.Y.C.R.R. Part 130, Judiciary Law Section 487, CPLR 2106, the Rules of Professional Conduct, and the Uniform Court System Interim Policy on Artificial Intelligence.

The existing enforcement record demonstrates that courts have sanctioned AI-related misconduct under pre-existing rules in every reported case to date. The bill introduces no civil penalty, no private right of action, no designated enforcer, and no safe harbor. Its definition encompasses tools – from spellcheck to optical character recognition to predictive coding – that the New York City Bar, the NYSBA task force, and the Office of Court Administration have all distinguished from the generative systems that produced the fabrications in Mata and its progeny. And it would operate alongside, rather than integrate with, the Office of Court Administration Interim Policy that addressed the same subject after 18 months of deliberation.

For practitioners, the immediate takeaway is that the duty of candor to the tribunal – codified in Rule 3.3, reinforced by Part 130, and illuminated by New York City Bar Formal Opinion 2024-5 and ABA Formal Opinion 512 – applies to AI-assisted work product regardless of whether A.8546 is enacted. Lawyers should independently verify all AI-generated citations and factual assertions, document their verification processes, and monitor developments in both the legislative and administrative frameworks as they evolve.

---

Alexander Paykin is the managing director and owner of The Law Office of Alexander Paykin, P.C., a boutique commercial and real estate litigation and complex transactions firm. He serves as chair of the NYSBA Committee on Technology and the Legal Profession, and as a member of the NYSBA Committee on Civil Practice Law and Rules, the NYSBA Law, Youth & Citizenship Committee, and the NYSBA Law Practice Management Committee. He sits on multiple ABA Law Practice Divisions. He regularly teaches CLEs for the ABA, NYSBA, and other state and local bar associations on technology and the practice of law. He speaks English and Russian.

The analysis in this article is the author’s own and not that of any committee, organization, or client.

Endnotes

[1] Mata v. Avianca, Inc., 678 F. Supp. 3d 443 (S.D.N.Y. 2023) (Castel, J.); see also Mata v. Avianca, Inc., Wikipedia, https://en.wikipedia.org/wiki/Mata_v._Avianca,_Inc.

[2] Park v. Kim, 91 F.4th 610 (2d Cir. 2024); see Eugene Volokh, Second Circuit Refers Lawyer for Disciplinary Proceedings Based on AI-Hallucinated Case in Brief, Reason.com (Jan. 30, 2024).

[3] Wadsworth v. Walmart Inc., 348 F.R.D. 489 (D. Wyo. 2025); see Bob Ambrogi, Federal Judge Sanctions Morgan & Morgan Attorneys for AI-Generated Fake Cases in Court Filing, LawSites (Feb. 25, 2025).

[4] Deutsche Bank Nat’l Trust Co. v. LeTennier, 2026 N.Y. Slip Op. 00040 (3d Dep’t. Jan. 8, 2026); see Jonathan Freiberger, Appellate Division, Third Department, Issues Monetary Sanctions against Attorney for Misuse of GenAI in the ‘First Appellate Level Case In New York’ To Do So, Freiberger Haber LLP (March 29, 2026), https://www.fhnylaw.com/appellate-division-third-department-issues-monetary-sanctions-against-attorney-for-misuse-of-genai.

[5] Damien Charlotin, AI Hallucination Cases Database, https://www.damiencharlotin.com/hallucinations/.

[6] N.Y. Assembly Bill A.8546, 2025-2026 Reg. Sess. (Lavine), https://www.nysenate.gov/legislation/bills/2025/A8546; N.Y. Senate Bill S.9794, 2025–2026 Reg. Sess. (Sepúlveda); N.Y. Senate Bill S.2698, 2025-2026 Reg. Sess. (Hoylman-Sigal), https://www.nysenate.gov/legislation/bills/2025/S2698.

[7] Carolyn Elefant, Public Comment on Proposed 5th Circuit Rule, in Public Comments on Proposed 5th Circuit A.I. Rule, E-Discovery LLC.

[8] 22 N.Y.C.R.R. § 130-1.1a.

[9] 22 N.Y.C.R.R. §§ 130-1.1(c), 130-1.2; see also Roy D. Simon, Stiffer Sanctions for Frivolous Litigation Conduct: Amended Part 130 Now in Effect, N.Y. Legal Ethics Rep, reprint from NYPRR, April 1998, ewyorklegalethics.com/stiffer-sanctions-for-frivolous-litigation-conduct-amended-part-130-now-in-effect/.

[10] Amalfitano v. Rosenberg, 12 N.Y.3d 8, 14 (2009).

[11] N.Y. Judiciary Law § 487; see USI Affinity, ‘Fraud on the Court’ and the Threat of Treble Damages: The Rise of Judiciary Law § 487 Claims Against Attorneys.

[12] N.Y.C. Bar Ass’n, Formal Op. 2024-5 (Aug. 7, 2024) (The Use of Generative AI in the Practice of Law).

[13] N.Y. State Bar Ass’n, Report and Recommendations of the Task Force on Artificial Intelligence (Apr. 6, 2024); see N.Y. State Bar Ass’n, NY State Bar Association Joins Florida and California in AI Ethics Guidance Suggests, Nat’l L. Rev.

[14] People v. Crabill, No. 23PDJ067 (Colo. O.P.D.J. Nov. 22, 2023); see Eugene Volokh, 90-Day Suspension of Colorado Lawyer Who Filed ChatGPT-Written Motion with Hallucinated Cases, Reason.com (Nov. 23, 2023).

[15] United States v. Hayes, 763 F. Supp. 3d 1054 (E.D. Cal. 2025); see Magistrate Judge Suggests Citations Were AI Fakes, Sanctions Federal Defender, Mealey’s (Jan. 24, 2025), https://www.mealeys.com/mealeys/articles/2288827.

[16] Da Silva Moore v. Publicis Groupe, 287 F.R.D. 182 (S.D.N.Y. 2012).

[17] N.Y.C. Bar Ass’n, Formal Op. 2024-5 (Aug. 7, 2024), at n.21.

[18] Grayned v. City of Rockford, 408 U.S. 104, 108–09 (1972); FCC v. Fox Television Stations, Inc., 567 U.S. 239, 253 (2012).

[19] Standing Order, Hon. Brantley Starr (N.D. Tex. May 30, 2023); see Shannon Capone Kirk et al, Judges Guide Attorneys on AI Pitfalls With Standing Orders, Ropes & Gray LLP (Aug. 2023), https://www.ropesgray.com/en/insights/alerts/2023/08/judges-guide-attorneys-on-ai-pitfalls-with-standing-orders.

[20]U.S. Court of Appeals for the 5th Circuit, Statement on Proposed Rule (June 12, 2024); see Ralph Losey, Fifth Circuit Court of Appeals Decides Not To Adopt Its Proposed Rule to Regulate the Use of Generative AI, e-Discovery Team (June 13, 2024) https://e-discoveryteam.com/2024/06/13/fifth-circuit-court-of-appeals-decides-not-to-adopt-its-proposed-rule-to-regulate-the-use-of-generative-ai/; Josh Blackman, 5th Circuit Seeks Comment on Proposed AI Rule, Reason (Nov. 29, 2023), https://reason.com/volokh/2023/11/29/5th-circuit-seeks-comment-on-proposed-ai-rule/.

[21] Public Comments on Proposed 5th Circuit A.I. Rule, E-Discovery LLC (collecting comments).

[22] Federal Court of Canada, Notice to the Parties and the Profession on the Use of Artificial Intelligence in Court Proceedings (Dec. 20, 2023; rev. May 2024).

[23] R (Ayinde) v. London Borough of Haringey & Al-Haroun v. Qatar Nat’l Bank QPSC, [2025] EWHC 1383 (Admin); see, Solicitors’ Negligence & AI: Ayinde v Haringey & Al-Harounv QNB Judgment Analysis, Carruthers Law (June 10, 2025), https://www.carruthers-law.co.uk/news/solicitors-negligence-ai-ayinde-alharoun-judgment. Also see Matthew Hermann et al, AI Hallucinations Hit the High Court, DAC Beachcroft (June 26, 2025), https://www.dacbeachcroft.com/en/What-we-think/AI-hallucinations-hit-the-high-court.

[24] Zhang v. Chen, 2024 BCSC 285; see, Recent BC Case Highlights Dangers of Generative AI, Canadian Lawyers Insurance Association, (March 1), https://www.clia.ca/not-so-risky-business/recent-bc-case-highlights-dangers-of-generative-ai.

[25] N.Y. State Unified Court System, Interim Policy on the Use of Artificial Intelligence (eff. Oct. 10, 2025); Greenberg Traurig LLP, Navigating AI Disclosure Rules in New York Courts (Nov. 2025).

[26] Hickman v. Taylor, 329 U.S. 495 (1947); N.Y. CPLR 3101(c).