May 13, 2026 – Artificial Intelligence (AI): It’s the tool everyone’s talking about, fretting over, and using. From doctors using medical scribes to businesses utilizing AI customer service agents to Hollywood investing in AI actors, AI and its potential capabilities are a main topic of discussion in boardrooms all across the world.
The legal profession is no different – in-house lawyers, law firms, and even courts are incorporating AI into their operations. Despite widespread adoption of AI tools in the legal industry, there are very few rules or regulations that govern how AI should and can be used.
In Wisconsin, there are no statutes or state rules that address the use of AI, much less the ethical use of AI in the legal profession. Due to this, courts are largely left without guidance on how to appropriately handle a litigant’s use of AI.
But is a new rule required just because a new technology is being used by the legal system? Lawyers are already governed by ethical rules about representations to the court, and Wis. Stat. § 802.05(2) addresses representations made by attorneys or self-represented parties in court filings. However, whenever new and innovative technology appears, the knee-jerk reaction is to implement a rule – even when there is only surface-level understanding regarding the technology.
As this article discusses, adopting the recently proposed AI rule (or other similar rules) would not only quickly become obsolete, but could lead to acrimonious litigation based on the challenges of defining generative artificial intelligence (GAI) and the systems that use it.
The Proposed Rule
On Feb. 19, 2026, a nonlawyer party submitted a proposed rule petition to the Wisconsin Supreme Court, requesting adoption of a specific rule to govern the use of GAI in court filings.[1]
Under the proposed rule, all litigants who use GAI must: 1) certify that GAI was used; 2) identify the specific tool or system used; and 3) provide a brief description of the purposes for which the tool was used.[2] The memorandum in support of the petition claims that the proposed rule will provide clear guidance on the use of artificial intelligence in Wisconsin Courts.[3]
Further, a supporting memorandum asserts that the proposed rule would clarify that existing ethical and professional responsibility rules apply to the use of AI.[4]
At first glance, it appears that the proposed rule could assist courts and litigants in navigating how to ethically use GAI; however, as the proposed rule and memorandum acknowledge, there are already rules in place related to a signer’s representations to the court and lawyers’ ethical obligations.[5] As such, a specific rule to regulate AI use is not needed unless there is another practical effect of the proposed rule.
The proposed rule’s primary purpose is to require litigants to disclose their use of GAI. It accomplishes this by requiring litigants to certify what, if any, GAI tools the litigant used in the creation of a document that is filed with the court.[6]
The proposed rule also requires lawyers, self-represented parties, and courts to review any content generated by GAI for accuracy.[7] The memorandum argues that the certification promotes transparency, accountability, and conscious verification of AI-assisted content without creating undue burden or discouraging beneficial uses of AI technology.[8]
Indeed, the memorandum claims that the proposed rule’s certification requirement makes clear that “lawyers, judges, and self-represented litigants remain personally responsible for ensuring the accuracy of all legal citations and authorities, regardless of how they were generated.”[9]
Questions Unanswered
Unfortunately, the proposed rule leaves too many questions unanswered. For example, the memorandum claims that the scope of the proposed rule defines the rule’s applicability to lawyers, judges, court staff, and self-represented litigants.[10]
However, the proposed rule’s statement of scope leads to a different conclusion. Proposed rule 76.01(a)(1) specifically states that “[t]his rule applies to a lawyer’s use of generative artificial intelligence tools” but does not state that the rule applies to self-represented litigants.
Instead, subsection (a)(3) states that self-represented litigants who use GAI “remain fully responsible for all documents and arguments they submit and are subject to all applicable statutes, rules of procedure, and rules of court concerning frivolous filings, false statements, or abusive litigation.”[11] Despite the exclusion of “this rule applies to” language in subsection (a)(3), the proposed rule imposes disclosure duties upon self-represented litigants.
These inconsistencies will only lead to confusion as the use of GAI increases. Likewise, there are several sections that discuss the courts; however, it does not appear that the proposed rule is meant to apply to courts because the courts are not required to file a certification for GAI use.
Instead, the proposed rule requires that courts continue to comply with the judicial code of conduct.[12] This begs the question of why the court is mentioned in the proposed rule at all. Ultimately, the lack of clarity in the proposed rule’s scope will cause confusion regarding to whom it applies and what consequences may follow for a self-represented party’s or court’s failure to abide by the proposed rule. [13]
Similarly – and this is a problem in many discussions about the use of “AI” – the definition of “generative artificial intelligence tool” is unclear, contradictory, and will likely open the door to complex sanctions litigation. Under the proposed rule, GAI tool is defined as “computer software or services capable of generating new content, including text, images, audio, or code, in response to user prompts.”[14] However, the definition “does not apply to the routine use of non-generative tools such as word-processing software, spell-checkers, citation-formatting tools, or traditional legal research databases that do not generate substantive content.”
The definition does not define “computer software or services,” makes assumptions of people’s knowledge of technology, and fails to take into account how rapidly AI technology evolves. For example, if a user puts a search into Google and receives an AI-generated summary, would using the summary require disclosure? Google is a search engine, but the text of the AI overview is generated using Google’s knowledge of a subject obtained from Gemini’s large language model[15] (LLM).[16]
Using just the AI summary could require disclosure if a litigant determines that Gemini is a “computer software or service[] capable of generating new content.” However, the technical expertise required to make this determination would most likely be beyond that of the average litigant.
Taking this example a step further, if a user clicks on the cited material from the AI-generated summary to the source (website), would a user be required to certify Google as a GAI tool? The mental gymnastics required to answer this question would not only be burdensome, but the answer would likely change depending on how each court interprets the definition of GAI tool. Ultimately, this lack of clarity could lead to potentially unnecessary sanctions litigation for failing to disclose what, if any, GAI tools were used to draft a document.[17]
Further, the GAI definition excludes “word-processing software” from mandatory disclosure. Based on the definition, a party would not be required to disclose using Microsoft Word. But such an exclusion fails to take into account that Microsoft Word with Copilot has GAI capabilities.[18] According to Microsoft, Copilot is a conversational AI-powered assistant that can help with document creation.[19] In other words, Microsoft Word with Copilot uses GAI to assist with document creation. If a party uses Microsoft Word with Copilot, would a certification for Microsoft Word be required? Confusion over the definition of a “GAI tool” will cause both underreporting or overreporting, which could then lead to more litigation over either the failure to disclose a GAI tool or discovery requests for prompts used with the disclosed GAI tool.
Likewise, the definition excludes “traditional legal research databases that do not generate substantive content.”[20] While the rule seems to plainly require disclosure if a party uses a legal research database’s GAI drafting tool, many legal research databases such as LexisNexis[21] and Westlaw[22] automatically utilize GAI to create content such as case summaries and headnotes.[23]
Does reliance on an AI-generated headnote constitute the use of a GAI tool for “substantive content”? Asked differently, absent a definition of “substantive content,” how would a party know if the AI-generated content was “substantive”? Once again, the exclusion causes confusion as to whether a party has to report using GAI.
Enforcement Authority
In addition to the unanswered questions it raises, the proposed rule does not provide any enforcement authority to the court (outside of its own authority to sanction parties) nor does it outline any guidance as to what the potential consequences or penalties could be for parties who fail to properly disclose GAI use. This lack of guidance could cause disparate consequences for litigants across the state where one court may sanction one party thousands of dollars while another court may only admonish the offending party.
But, as the memorandum and proposed rule themselves acknowledge, the proposed rule only reiterates preexisting obligations by which litigants and the courts must abide.[24] That acknowledgement in mind, rather than adopting a rule that governs GAI use, the more practical approach is to rely on pre-established rules of conduct, which are already flexible enough to adapt to litigants delving into the ever-changing GAI landscape.
For example, the Illinois Supreme Court recently adopted an AI policy effective on Jan. 1, 2025, that allows the court and litigants to use AI tools in legal proceedings provided such use complies with the Illinois Rules of Professional Conduct and Code of Judicial Conduct.[25]
According to the Illinois AI Policy, the policy embraces the advancement of technology while reminding litigants of their ethical obligations regarding work product.[26] The Illinois AI Policy is a practical solution to regulating the use of AI because AI is constantly advancing and changing.
Defining AI in a rule today may not capture what AI is next year. This constant advancement of technology could create issues where the rule would have to be regularly reviewed and amended to account for technological changes. The Illinois approach of relying on pre-established rules reduces the potential of immediate obsolescence or the need for repeated amendments to a rule due to technological advancements. Additionally, this approach does not create new rules but reminds litigants and the courts of their existing obligations.
Indeed, the commonsense approach of relying on pre-established rules is illustrated in Wisconsin Formal Ethics Opinion EF-15-01: Ethical Obligations of Attorneys Using Cloud Computing.[27] The opinion acknowledges that “[i]t is impossible to provide specific requirements for reasonable efforts because lawyers’ ethical duties are continually evolving as technology changes. Specific requirements would soon become obsolete.”[28]
Additionally, “specific requirements would do little to assist the exercise of professional judgment” because “[l]awyers must exercise their professional judgment in adopting specific cloud-based services, just as they do when choosing and supervising other types of service providers.”[29] Instead, the Ethics Committee reminds attorneys that when deciding whether to implement cloud computing, the rules of competence, communication and confidentiality still apply. As with cloud computing, GAI evolves at such a rapid pace that a specific rule would quickly become obsolete. Recreating the wheel to address every new type of technology would ultimately be a futile endeavor. As the proposed rule itself acknowledges, everyone who appears before a court in Wisconsin is already bound by multiple duties and ethical obligations. The development of a new technology does not change those obligations.
Recommendations
If Wisconsin decides not to follow Illinois’s lead, then any proposed rule should address the questions raised above about the proposed rule (although these are just some of the potential issues with the proposed rule), which, if left unanswered, will lead to overreporting, underreporting, potential litigation over the required certification and the use of GAI.
As discussed above, failure to properly understand the technology that the rule is attempting to regulate will only lead to confusion and potential unintentional violations of the rule.
However, as many regulatory groups have discovered, developing a precise definition of AI is difficult.[30] This is because the decision to use human based language to define GAI may not fully encapsulate all GAI systems but using a technical definition of GAI may fail to provide a usable definition to the common user.
Because of the complexities in defining GAI, if a rule must be adopted, the rule should be based on the recommendations of a committee of legal practitioners and technological experts who can more appropriately define GAI and the potential rule’s scope. That way, any proposed rule will more accurately capture both the constant evolution of GAI and its use in the legal profession. But, for now, until the above questions (and any others) are ironed out, adoption of the proposed rule is premature and could create more problems than it solves.
Endnotes
[1] Petition 26-02: The Matter of Creation of Supreme Court Chapter 76 (Use of Artificial Intelligence Tools), February 19, 2026, https://www.wicourts.gov/supreme/docs/2602petition.pdf. ^
[2] Id. at 2 (Proposed SCR 76.01(c)(2)). ^
[3] Memorandum: In the Petition to Create SCR Chapter 76 (Use of Artificial Intelligence Tools), https://www.wicourts.gov/supreme/docs/2602memo.pdf at 1. ^
[4] Id. ^
[5] See SCR 20:3.3 (lawyers shall not knowingly make a false statement of fact or law to a tribunal); see also Wis. Stat. 802.05(2) (certification of representations made to the court by signing, filing, submitting or other document). ^
[6] Petition at 2 (76.01(c)). ^
[7] Petition at 2 (SCR 76.01(c) and (d)). ^
[8] Memorandum at 12. ^
[9] Memorandum at 5. ^
[10] Id. ^
[11] Petition at 1. ^
[12] See Petition at 2 (76.01(c) (only requiring lawyers and self represented litigants to verify and disclose GAI use)). ^
[13] For example, it is unclear what “personally responsible” means under the proposed rule. See Petition at 3 (SCR 76.01(c)(3) (“A self-represented litigant remains personally responsible for the content of all submissions.”)) (emphasis added); see also Petition at 3 (SCR 76.01(d) (“[courts] that use generative artificial intelligence tools in connection with the drafting of orders, opinions, or other court documents remain personally responsible for the content of such documents…”)) (emphasis added). Does this mean that the self-represented litigant and the court are subject to personal sanctions? ^
[14] Petition at 2 (SCR 76.01(a)(3)). ^
[15] Cole Stryker, What are large language models (LLMs)?, https://www.ibm.com/think/topics/large-language-models#692473873, (last visited April 21, 2026) (“Large language models (LLMs) are a category of deep learning models trained on immense amounts of data, making them capable of understanding and generating natural language and other types of content to perform a wide range of tasks.”) ^
[16] What are Google AI overviews and how do they work?, https://www.botify.com/insight/what-are-google-ai-overviews, (last visited April 21, 2026). ^
[17] See Wis. Stat. 802.05(3) (procedure for sanctions for violating Wis. Stat. 802.05(2)). ^
[18] What is a copilot? Microsoft Copilot, https://www.microsoft.com/en-us/microsoft-copilot/copilot-101/what-is-copilot (last visited April 9, 2026). ^
[19] Id. ^
[20] There is no definition of “traditional legal research databases” or “substantive content”. ^
[21] Frequently Asked Questions, LexisNexis, https://www.lexisnexis.com/en-us/products/lexis-plus.page#faqs (last visited April 9, 2026). ^
[22] Westlaw Edge with AI-Assisted Research, Thomson Reuters, https://legal.thomsonreuters.com/en/products/westlaw-edge (last visited April 9, 2026). ^
[23] 6 Enhancements to Lexis AI that Improve Legal Research and Drafting Workflow, LexisNexis, https://www.lexisnexis.com/community/insights/legal/b/thought-leadership/posts/6-enhancements-to-lexis-ai-that-improve-legal-research-and-drafting-workflow (last visited April 20, 2026) (“Lexis+ AI now generates headnotes for every case, enabling users to better understand the points of law for each decision and to easily locate similar cases by related points of law.”). ^
[24] See Wis. Stat. 802.05(2). ^
[25] Illinois Supreme Court Policy on Artificial Intelligence, https://ilcourtsaudio.blob.core.windows.net/antilles-resources/resources/e43964ab-8874-4b7a-be4e-63af019cb6f7/Illinois%20Supreme%20Court%20AI%20Policy.pdf (last visited April 9, 2026). ^
[26] Id., at 2 (“This policy reflects the Illinois Supreme Court’s commitment to upholding foundational principles while exploring the potential benefits of new AI technologies in a dynamic landscape.”). ^
[27] WI State Bar Prof’l Ethics Comm., Formal Op. EF-15-01: Ethical Obligations of Attorneys Using Cloud Computing, Amended September 8, 2017, https://www.wisbar.org/formembers/ethics/Ethics%20Opinions/EF-15-01%20Cloud%20Computing%20Amended.pdf. ^
[28] Id. at 10. ^
[29] Id. at 11. ^
[30] See Matt O'Shaughnessy, One of the Biggest Problems in Regulating AI Is Agreeing on a Definition, Carnegie Endowment for International Peace (Oct. 6, 2022), https://carnegieendowment.org/posts/2022/10/one-of-the-biggest-problems-in-regulating-ai-is-agreeing-on-a-definition. ^
