As a practicing attorney, working under a public defense contract, the uptick in clients coming to me with reems of legal documents generated by AI has started to approach overwhelming. Thus far, citing court decisions and instances where attorneys are sanctioned for using AI-generated material with no actual foundational legal basis (still common in AI generated material), but presenting it to the court as if it does have legal foundation, has been enough of a bulwark from becoming a full-time legal researcher for my clients who have stars in their eyes due to AI-generated fantasy outcomes. That said, these documents put me on my back foot out the gate, and erodes my role as an attorney with real life experience, who is making recommendations and giving sound legal counsel on potential real-life outcomes, based on evidence and real-world legal knowledge. Further, and to make a distinction here, in my experience, AI-lawyering is distinctly different from google-lawyering. Google-lawyering is a phenomenon weʼve been dealing with for decades now. But google-lawyering simply provides law without context. AI gives context based, fact-specific legal argument, and generates material with those fact-specific circumstances included, which appears at a glance to have the weight of legal authority. These authoritative looking documents in the hands of uninformed and unskilled clients creates very real rifts between attorneys and their clients, as the attorney needs to explain why the client is wrong. When in reality, weʼre essentially fighting their pocket attorney: AI. And these types of rifts are, frankly, typically what cause clients to want to proceed pro se. My fear is while this new rule protects the court to a degree, it keeps the door *too* open to AI-generated material being admissible to the court. And further, it puts the attorney in an attorney-client relationship on the hook, if that AI-generated material is found lacking. In more practical terms: I need more ammunition to tell my client why I wonʼt be submitting their AI-generated motion, not a rule that allows them to tell me its admissible, I simply need to review it first. Then the other side of the coin is, the nature of AI is that it presumably continues to get better and better. It is designed to "learn," and with each of its iterations, it creates something better than the last. At the same time, it does not (to my knowledge) currently have a feedback loop which includes the court, the courtʼs rulings or any kind of actual outcome-based result to learn off of. So, while AI-generated material is still arguably in its neanderthal phase, and many times easily dismissible with a basic, cursory readthrough, it is an open question as to how long that reality will last. I understand an outright ban on AI material, while that is what I would like to see, is not pragmatic or practical - especially from an enforcement perspective. But at the same time, there is a tidal wave of legal documents on the horizon baring down on all of us in the justice system. AI has undeniably made the generation of legal documents and motions readily accessible to anyone with a library card, and anyone with a library card is starting to realize this gives them a way to argue for what they want, even if it is outlandish and fantastical. While addressing the AI issue is undeniably important, this rule as written unfortunately reads as a temporary band-aid on a situation that could get out of hand very soon, and very quickly. It simply feels like a burden shift that will only do so much: it makes attorneys the AI gatekeepers. And seems to solely focus on the issue of pro se defendants generating their own AI documents, or attorneys generating them themselves. However, when attorneys in attorney-client relationships have to be the ones to tell a client why they wonʼt get what they want, why they wonʼt be proceeding with their AI-generated motion, when AI is telling them they can get what they want, who do you think the client wants to listen to? Lastly, for a state and jurisdiction dealing with a public defense crisis, my other fear is this rule will only exacerbate that problem, as it would create even more of a per-client workload. Simply put, there is a big difference between a blanket "I cannot submit AI generated material to the court, as the court simply believes it is too wrong too often" and "let me tell you why your AI motion is wrong based on my knowledge and the research Iʼve had done in regard to your AI generated motion..." Thank you for your time, and good luck addressing this very important AI issue.

I am writing regarding the proposed local rule, LGR 30.1, addressing the use of artificial intelligence in court submissions. I am attaching the recent WSBA advisory opinion on this issue and respectfully ask that this email be provided to the judges considering the proposed rule.

I understand the concern that likely prompted the proposal. Courts are understandably concerned about filings that contain fabricated cases, unsupported factual claims, unreliable AI-generated material, or lengthy filings that are not grounded in law. Those are real concerns. But the proposed rule does not solve that problem, and it creates unnecessary issues for lawyers who are already bound by existing rules and ethical obligations.

The WSBA advisory opinion already addresses the actual issue. Attorneys remain responsible for their filings. They must verify legal authority, confirm factual assertions, protect client confidences, supervise the use of technology, and comply with their duties of competence, diligence, candor, and confidentiality. Likewise, CRLJ 11 already requires an attorney or party who signs a filing to certify that it has been read, is well grounded in fact, is warranted by existing law or a good-faith argument for changing the law, and is not filed for an improper purpose.

The proposed rule is therefore unnecessary as applied to pleadings, motions, briefs, and other legal submissions. If a lawyer files a brief containing fake cases or false statements, the issue is not whether AI was used. The issue is that the lawyer filed material that violates existing rules. CRLJ 11 already addresses that. The Rules of Professional Conduct already address that. A separate AI-disclosure rule adds burden without adding meaningful protection.

The most concerning part of the proposed rule is subsection (c), which requires disclosure of AI assistance in the “creation or editing of any document or evidence submitted to the court,” including “a general description of the AI technology used and its role in the preparation of the materials.” That requirement is overbroad and intrudes into attorney work product.

How an attorney researches, drafts, edits, proofreads, or organizes a filing is not ordinarily a matter for disclosure to the court. Modern legal research and drafting tools increasingly include AI or AI-like features. Lexis+ AI, Westlaw, Microsoft Word, Adobe, Grammarly, transcription platforms, document review platforms, and other tools may use machine learning, natural language processing, or generative AI in ways that are not always obvious to the user. Requiring counsel to disclose the “technology used” and its “role” in preparing a filing invites uncertainty, satellite disputes, and unnecessary certifications regarding internal attorney work.

For example, if an attorney uses Lexis+ AI to identify potentially relevant cases, the attorney remains responsible for reading those cases and verifying the law. The research method is attorney work product. The court does not need to know whether counsel used Lexis+ AI, Westlaw, Google Scholar, a treatise, a law clerk, a colleague, or a prior brief to conduct research. The relevant question is whether the final filing is accurate, supported, and compliant with CRLJ 11.

The same is true for editing. If an attorney uses an AI-assisted editing tool to improve grammar, shorten sentences, identify unclear wording, or proofread a draft, that should not trigger a disclosure obligation. It serves no legitimate purpose for the court to require lawyers to certify the internal tools used to prepare a document. What matters is the final submission, not the drafting path used to get there.

Evidence is different. If AI is used to create, alter, enhance, reconstruct, translate, summarize, or analyze evidence offered to the court, disclosure may be appropriate because the AI process may bear on authentication, foundation, admissibility, reliability, or weight. But those concerns are already addressed through the evidence rules, discovery rules, expert-disclosure rules, and ordinary motion practice. A narrow evidence-focused rule might be understandable. A broad rule requiring disclosure of ordinary legal research, drafting, proofreading, or editing is not.

The proposed rule is also unlikely to address the problem of pro se AI-generated filings. The litigants most likely to file AI-generated nonsense are unlikely to read and comply with a new local rule requiring AI disclosure. Courts already have tools to address improper filings, including CRLJ 11, page limits, orders to show cause, striking improper materials, and other existing remedies.

For those reasons, I respectfully object to the proposed rule as unnecessary, overbroad, and intrusive into attorney work product. The better approach is to rely on the WSBA advisory opinion, CRLJ 11, the Rules of Professional Conduct, and the existing evidence and discovery rules. At most, any local rule should be narrowly limited to AI-created or AI-altered evidence where authenticity, foundation, admissibility, reliability, or weight are genuinely at issue.

Thank you for passing this on to the judges considering the proposal.

WA Bar Association Advisory Opinion 2025-05

Clarification on whether disclosure of any use of AI (for research for example) is needed, or just if AI is used in the actual creation of the document/filing. Does "prepared," "creation," and "editing" refer to just the format or writing/language used, or any usage of AI? 

I understand the intent of the rule but it seems like it might be a bit vague and overbroad. Lexis and Westlaw both have specific AI assisted research options now. In addition, the mere use of legal research platforms where a question is asked using plain language or using search terms probably qualifies as use of AI I also think it is not uncommon to start legal research with an AI query on Lexis or Westlaw which may give a proposed answer and sources. If you are a good lawyer that is the start not the end of your research. To an extent I think this rule would require the filing of a certificate whenever a brief is filed explaining how Lexis/ Westlaw were used to research the law.

I write in opposition to Proposed LGR 30.1. I have practiced in Clark County for over a decade and have read the proposal carefully. The rule will not accomplish what it sets out to accomplish, and it will load paperwork onto the people least likely to need it. The conduct that drives proposals like this one, fabricated case citations and unreviewed authority, is already prohibited. CRLJ 11 requires every signer of a pleading or motion to certify, after reasonable inquiry, that the filing is grounded in fact and warranted by existing law. RPC 1.1 requires competence. RPC 3.3 imposes an absolute duty of candor toward the tribunal. RPC 5.1 and 5.3 require supervision. RPC 8.4 prohibits dishonesty. The attorneys in Mata v. Avianca, Inc., 678 F. Supp. 3d 443 (S.D.N.Y. 2023), were sanctioned under existing Rule 11. They did not need a disclosure rule. They needed to read what they signed. Every Washington court already has the authority to sanction a lawyer who files fabricated authority, no matter whether the source is ChatGPT, an unsupervised paralegal, or the lawyer’s own carelessness. The larger problem is the rule’s definition of AI. As written, it captures essentially every digital tool a Washington lawyer now uses. Westlaw Precision, CoCounsel, Lexis+ AI, Lexis Protege, Bloomberg Law, Fastcase, and vLex all rely on machine learning to surface and rank authority. The current version of Microsoft Word ships with Copilot, and Microsoft Editor goes well past spell check; it suggests substantive rewrites and tone changes. Outlook offers Copilot drafting. Gmail has Smart Compose. Google Docs has integrated generative AI. Grammarly is itself an AI product. Adobe Acrobat uses AI for OCR and document review. Voice-to-text dictation on modern phones and in Dragon uses neural networks. Google Translate and DeepL, which I and most criminal and immigration practitioners rely on routinely, are AI products. Clio, MyCase, PracticePanther, and Smokeball all advertise AI features for time entry, intake, and document automation. Read literally, the rule demands a separate certification every time any of those tools touches a filing. The lawyer who tries to comply in good faith ends up disclosing Microsoft Editor on every routine motion. The lawyer who shrugs quietly violates the rule on every filing. The carve-out for spelling and grammar checking does not save it, either. Modern grammar engines, including Microsoft Editor, Grammarly, and the tools built into Google Docs and Apple’s system writing features, are themselves AI. The line between “checking grammar” and “rewriting prose” exists only in marketing copy. The harder problem is enforcement. The fabricated-citation issue in courts around the country has been driven largely by pro se litigants using free chatbots without understanding how they hallucinate. The proposed rule will not reach those filings. Pro se litigants do not read local rules. This court hears substantial volumes of small claims, anti-harassment, infraction, and other matters in which one or both parties are unrepresented, and those parties will not file an AI certification no matter what the rule says. Many of them have no idea what AI is. And realistically, this court is not going to strike a pro se anti-harassment petition or a small claims complaint solely for the absence of a certification, particularly where the cited authority is in fact accurate. The burden of the rule will fall almost entirely on lawyers, who are already bound by the RPC and CRLJ 11 and who are not the population causing the problem. Once a rule like this is on the books it also becomes a tactical instrument. Opposing counsel will move to strike on the theory that a brief was AI-assisted and not disclosed. Discovery will probe AI use in drafting. Hearings will be consumed by arguments over whether a Westlaw Precision search triggered the certification or whether a Copilot edit was substantial enough to count. None of that decides anyone’s case. If the court is determined to act in this area, a narrower path is available. Require a single standing certification from counsel that all cited authority has been independently verified to exist and to stand for the proposition cited. Post a plain-language warning to pro se litigants at the clerk’s counter and in the e-filing portal, where they will actually see it. And sunset any rule that is adopted at twelve or eighteen months for review, because the technology is moving faster than rule-making. That approach reaches the actual harm without dragging every modern legal tool into a paperwork regime. The existing rules already do the work this proposal was designed to do. I urge the court to decline to adopt LGR 30.1, or, if some rule is to be adopted, to substantially narrow it first. Thank you for considering this comment.