New York’s Part 161 v. California’s SB 574: Two very different approaches to AI use in court.
Opposite sides of the country, opposite sides of the AI line in the sand?
In 2027, litigators in New York and California will be following two very different rules when using generative AI tools. I draft appeals and complex motions for attorneys and law firms in New York and California (as well as other states), which means I need to understand what each state requires from its lawyers’ work product. So I took a deep dive to understand the differences – and because my “writing is not writing at all, it is thinking,” I’m writing this post to both cement my knowledge and share it with you.
Scroll all the way down for a chart with a side-by-side comparison of the two states’ rules (and of course I used generative AI to create it after reading both sets of rules first).
On June 1, 2026, the New York State Unified Court System put into effect 22 NYCRR Part 161 (Use of Artificial Intelligence Technology), its first statewide court rule governing AI in civil and criminal filings. Four months later, on September 30, 2026, California Governor Gavin Newsom signed Senate Bill 574 (SB 574, Chapter 858, Statutes of 2026), the first comprehensive statute in the country that regulates how attorneys (and arbitrators) use generative AI. SB 574 takes effect on January 1, 2027.
Both states agree on the core premise: AI is here to stay, lawyers are allowed to use it, and a lawyer is 100% responsible for fake citations or hallucinated law.
The states split — and dramatically so — on disclosure, scope, and how much the state wants to micromanage your workflow.
New York (well, at least the Chief Administrator of the Courts) says you do not need to tell the court you used AI because your signature already guarantees your work is accurate. Part 161 seems to be a reaction to the increasing rules in judges’ parts that required attorneys to file a certification with each filing that informs the court and adversaries what AI tools were used and for what.
California, by contrast, is requiring mandatory AI disclosure on every single court submission, codifying strict data-privacy rules for client inputs, and holding law firms jointly liable for unverified citations.
Here is where both rules stand, how they compare, and what they mean for your practice.
How does New York’s Part 161 regulate attorney use of AI?
Part 161 takes a pragmatic, judicial-administration approach. It applies to all courts of the Unified Court System across both civil and criminal cases (§161.1) and covers any "paper" — defined as a brief, memorandum, affidavit, affirmation, pleading, or other document prepared for submission to a court, while excluding materials proffered as evidence (§161.2(b)).
The centerpiece of the rule is §161.3 (Policy), which answers the mandatory-disclosure question with a clear "no":
"It is the policy of the Unified Court System that the use by attorneys and parties of artificial intelligence tools in preparing papers submitted to a court should not be prohibited, as long as such use is in accordance with the duties and responsibilities that apply to individuals who submit papers to a court. Since those duties and responsibilities already apply to all submissions, regardless of whether AI tools were used, attorneys and parties should not be required, upon submitting papers, to disclose to the court that they have used AI in the preparation of such papers." (emphasis added).
As an appellate attorney who has come across myriad wrong citations and flawed legal reasoning in trial level papers, I think Part 161 is logical and intellectually honest. Under existing New York rules — for example, 22 NYCRR §130-1.1 and §130-1.1a — every time a lawyer signs a brief or motion, she is already certifying that the paper contains no false material factual statements and no frivolous legal arguments. There is a premium on accuracy in the courts. The tool used by the lawyer isn't the issue, sloppy lawyering is.
But before you assume you never have to think about AI certifications in New York, beware a caveat: Under §161.4, an individual court (i.e., judge) may, in their discretion, implement a part rule governing AI use in court papers. The element of discretion is similar to the one in New York’s rule on Statements of Undisputed Material Facts under 22 NYCRR §202.8-g.
If the court chooses to adopt a part rule on AI use, §161.4 encourages them to adopt the Appendix A Model Rule, which provides:
"Accordingly, any attorney or party who uses an artificial intelligence tool, as defined in 22 NYCRR 161.2(a), in preparing any paper, as defined in 22 NYCRR 161.2(b), filed in or submitted to this court or served on another party in a case before this court is required to carefully review the paper and independently ensure that it contains no fabricated or fictitious cases, statutes, or other material. By signing such paper, an attorney or party certifies that such a review has been conducted and that the paper contains no such fabricated or fictitious content. If this court determines that this requirement has not been satisfied, such attorney or party may be subject to sanction or other remedial action."
Just like before the AI era, a lawyer’s signature on the paper is the certification that everything in the paper is accurate.
Despite the enactment of Part 161, some New York judges have still adopted individual part rules that require affirmative certifications identifying the generative AI program used and for which parts of the paper submitted. Always check and follow your judge’s part rules. Do not rely on Part 161 to pick a fight with your judge about AI disclosure. But I imagine judges who implement strict part rules will be guided toward compliance with Part 161’s permissive structure – because it sets a sensible baseline that is consistent with the duty of competence and candor that is already owed by lawyers to the New York courts.
How does California’s SB 574 regulate attorney use of AI?
No short court rule focused on litigation papers for California attorneys. Instead, the Legislature passed a sweeping statute. California SB 574 adds Section 6068.1 to the Business and Professions Code (the State Bar Act), amends Code of Civil Procedure (CCP) Section 128.7 (California’s equivalent of Federal Rule 11), and adds CCP Section 1282.1 for arbitrations.
Unlike New York’s rule, California’s statute isn't limited to court filings — it governs the practice of law across the board for California-licensed attorneys.
Here are five major rules in the new law:
No delegating legal judgment to AI (Bus. & Prof. Code §6068.1(a)(1)–(2)): The statute plainly states that "[a]n attorney shall not delegate the practice of law to generative artificial intelligence," while preserving existing duties of competence and diligence. The statute does not define “delegate” or give any indication of what that term means.
Mandatory disclosure on every court submission (Bus. & Prof. Code §6068.1(a)(3)(C)): In a 180-degree difference from New York, California requires an attorney who uses generative AI to "[d]isclose the use of generative artificial intelligence to the court for all documents submitted to the court," and to consider whether to disclose AI use in content provided to the public. The statute does not specify what that disclosure must look like or whether incidental AI features built into research platforms or word processors trigger it.
Strict confidentiality rules banning consumer AI tools for client data (Bus. & Prof. Code §6068.1(a)(3)(A), (b)(2)): Attorneys may not enter confidential, personally identifying, or other nonpublic information into a generative AI system unless access to those inputs is restricted to the attorney and authorized persons under confidentiality obligations. The law defines "personal identifying information" broadly to include names/addresses/phones of parties, victims, witnesses, and court personnel, dates of birth, SSNs, medical or psychiatric details, financial data, and sealed records. It’s not specified in the statute, but it seems like if a consumer AI tool retains inputs for training or allows vendor human review for abuse monitoring, putting client “personal identifying information” into the tool will run afoul of the law.
Verification across all legal work — even outside court (Bus. & Prof. Code §6068.1(a)(3)(B)): Attorneys must take "reasonable steps" to verify the accuracy of generative AI outputs (including all case and statutory citations) and correct any erroneous or hallucinated output "in any material used by the attorney." That means a hallucinated case in an internal research memo or client email is now a statutory violation enforceable by State Bar discipline, even if it never gets filed in court.
"Personal verification" of citations and joint law-firm liability (CCP §128.7(b)(2)(c)–(d)): SB 574 amends California’s sanctions statute to mandate that:
"A brief, pleading, motion, or any other paper filed in any court shall not contain any citations that an attorney responsible for submitting the pleading has not personally verified, including any citation provided by generative artificial intelligence."
The language "personally verified" is a throwback to the basics of lawyering: if a citation turns out to be wrong, a signing attorney cannot blame a junior associate, paralegal, or cite-checker. This has always been the case. The difference now is that AI tools are often generating the errors, even if the human lawyers also still make the same kinds of mistakes.
Notice also that the “personal verification” rule applies to any citation in a California court filing, whether generated by AI or not, and excludes discovery requests and responses (§ 128.7(g)). And while the 21-day safe harbor to withdraw or correct a filing still applies, "[a]bsent exceptional circumstances, a law firm shall be held jointly responsible for violations committed by its partners, associates, and employees." Cal. Code Civ. Proc. §128.7(c)(1), (d)(1).
My take as an appellate attorney: substance still beats technicalities.
When New York amended CPLR 2106 to let litigants submit unnotarized affirmations instead of notarized affidavits, I observed that California was the state that already had the simpler, more practical rule. On AI disclosure in the courts, the two states have traded places in my view (though I will follow the AI rules in every state court I write briefs for, no matter what they are).
I have made no secret of my view on procedurally nitpicky rules: I hate when technicalities distract from the substantive merits of an argument. California’s blanket requirement that lawyers disclose generative AI use on all documents submitted to the court departs too much from the reality of law practice. The law does not define what level of AI assistance triggers it, and it ignores the fact that even human lawyers make citation errors that lessen the quality of their work product. Generative AI is embedded into email, spell-checkers, Westlaw, Lexis, Word, Adobe; California’s rule forces lawyers to tack a boilerplate AI disclosure onto every brief while doing very little to improve the quality of legal analysis and writing.
New York’s rule tracks closer to the legal profession as it is. Whether you drafted a brief with a quill pen, a word processor, or the help of a large language model, your signature on the brief means you verified every case, every quote, and every factual assertion. When I draft a brief or motion for you, I verify the same when I send you the draft.
As I shared on my FAQs page and in my recent note on why trial lawyers shouldn't rely on AI to handle their own appeals, I use generative AI tools to augment — not automate — my work. I hate reading AI slop, and I never put it in my briefs. Even when an AI tool cites a real case, it frequently misses the nuance that is the difference between a win and a loss; it sometimes misses the procedural posture, overlooks a statute that superseded an older holding, or misses a subtle factual clue buried in a 2,000-page record that might be the key to a win.
Whether you are practicing under New York’s Part 161 or gearing up for California’s SB 574 on January 1, 2027, no certification block or software tool can replace solid lawyering: reading every case cited, building an argument from the ground up.
Need human expertise for your client’s brief or complex motion?
If you are a busy trial lawyer or small firm tired of wrestling with AI output — and you want an appellate attorney with two decades of experience who reads the entire record, reads every case cited, and sends you a rule-compliant, ready-to-file brief — get in touch with me at On Point Expertise to schedule a time to discuss your client’s appeal or complex motion. State and federal in New York, California, and courts beyond.
Sign up for the On Point Expertise newsletter to stay in touch.
New York Part 161 v. California SB 574: Side-by-Side Comparison
| Feature | New York (22 NYCRR Part 161) | California (SB 574 / Bus. & Prof. Code § 6068.1 & CCP § 128.7) |
|---|---|---|
| Governing Instrument | Court rule adopted by Chief Administrative Judge (22 NYCRR Part 161) | Legislative statute amending the State Bar Act and Code of Civil Procedure (Ch. 858, Stats. 2026) |
| Effective Date | June 1, 2026 | January 1, 2027 (signed Sept. 30, 2026) |
| Who & What Is Covered | Attorneys and self-represented parties submitting “papers” in all NY Unified Court System civil and criminal courts (excludes evidence) (§§ 161.1–161.2) | CA-licensed attorneys across all legal work (§ 6068.1), all attorneys filing papers in CA civil courts (CCP § 128.7), and arbitrators (CCP § 1282.1) |
| Definition of AI | Broad: any machine-based system making predictions, recommendations, or decisions (§ 161.2(a)) | Specific to Generative AI: systems generating derived synthetic content (text, images, video, audio) (§ 6068.1(b)(1)) |
| Mandatory Disclosure to the Court? | No statewide requirement (§ 161.3), BUT individual judges’ part rules might require it under § 161.4 | Yes, statewide by statute. Mandatory disclosure of GenAI use for all documents submitted to the court, plus consideration of disclosure for public content (§ 6068.1(a)(3)(C)) |
| Citation Verification Standard | Under the optional Appendix A Model Rule, signing certifies careful review and independent assurance of no fabricated citations or facts | Responsible attorney must personally verify every citation in any court filing (CCP § 128.7(b)(2)(A)) and take reasonable steps to verify all AI outputs in any work product (§ 6068.1(a)(3)(B)) |
| Client Confidentiality & PII Rules | Governed by existing NY Rules of Professional Conduct (Rule 1.6); not separately addressed in Part 161 | Express statutory ban on entering confidential or broadly defined PII into GenAI unless access is restricted to authorized, confidentiality-bound users (§ 6068.1(a)(3)(A), (b)(2)) |
| Sanctions & Firm Liability | Existing court sanctions under 22 NYCRR 130-1.1 / 130-1.1a, professional discipline, or denial of application under individual part rules (Appendix A) | CCP § 128.7 sanctions (21-day safe harbor), joint law-firm liability absent exceptional circumstances, no monetary sanctions on represented clients, plus State Bar discipline (CCP § 128.7(c)–(d)) |