Illinois Judge’s Rules for AI in Litigation May Preview Coming Rules Across the Country

In September 2025, Judge Sarah D. Smith of the Third Judicial Circuit in Madison County, Illinois, issued a standing order establishing detailed rules for attorneys’ use of artificial intelligence in civil litigation that could offer an early look at how courts may regulate lawyers’ use of AI in courts across the country. The order permits responsible use of AI but imposes requirements addressing pleadings, discovery, confidentiality, depositions, evidence, experts, and jury selection. 

Nearly a year later, in August 2026, Judge Smith issued an addendum addressing the other side of the issue: how the court itself uses AI. This is one of the first known judicial orders of its kind in the state. The addendum explains that AI may assist chambers with certain administrative and drafting functions, but that judicial decision-making remains solely with the judge. Together, the initial order and later addendum provide a useful preview of what courts may expect from lawyers using AI and explain their own use of the technology. AI is permitted, but human responsibility remains.

Judge Smith’s September 2025 standing order does not prohibit lawyers from using generative AI. It recognizes that AI may improve efficiency and productivity, but also identifies risks involving accuracy, authenticity, bias, confidentiality, and the integrity of court proceedings. The order expressly permits AI as a supplemental tool while requiring human oversight and legal judgment at every stage. 

Lawyers and self-represented litigants must personally verify the factual and legal accuracy of material generated with AI. The attorney signing a pleading remains responsible for its contents under Illinois Supreme Court Rule 137, regardless of whether AI, staff, or someone else helped prepare it. Inaccurate propositions of law, hallucinated cases, and “ghost citations” may result in sanctions. 

That approach is generally consistent with the statewide policy adopted by the Illinois Supreme Court effective January 1, 2025. The statewide policy authorizes responsible AI use, does not require disclosure of AI use in pleadings, and places ultimate responsibility for the final work product on the lawyer, judge, or litigant submitting it. 

The order goes well beyond pleadings 

The most notable aspect of Judge Smith’s initial order is its scope. It treats AI not merely as a legal-writing tool, but as technology that may affect virtually every part of a case. 

Discovery and document review 

Discovery responses prepared with AI must be verified for factual accuracy, legal sufficiency, and compliance with the Illinois Supreme Court Rules. Lawyers may not rely on AI-generated boilerplate, general objections, or speculative assertions. If AI or technology-assisted review is used for document review or interrogatory preparation, counsel must disclose the method upon request and provide adequate validation and human oversight. 

The order also requires disclosure when AI or an automated summarization tool helped answer interrogatories or summarize client records. The disclosure must be made to opposing counsel when the responses are served and identify the general category of tool used. The disclosure itself does not waive privilege, assuming the tool was used in a manner that maintains the privilege in the first place. 

Counsel must also file a signed certification confirming personal review of the responses, verification of their accuracy and completeness, and the application of independent legal judgment. Failure to comply, including reliance on erroneous or incomplete AI-generated responses, may result in sanctions under Illinois Supreme Court Rule 219. 

Confidentiality and privilege 

The order warns attorneys not to upload confidential, privileged, or protected health information into AI systems that store or transmit data externally unless adequate protections are in place. Failure to safeguard that information may result in sanctions and may be treated as a waiver of privilege. 

That requirement makes vendor selection and data handling more than an internal technology issue. Lawyers need to know whether a tool retains prompts, uses information for model training, transmits data to third parties, or stores information on an external server. 

Depositions and courtroom proceedings 

A witness is “absolutely prohibited” from using AI to generate, suggest, or supplement deposition answers. The order treats that conduct as impermissible coaching that interferes with the truth-seeking function of a deposition and may support sanctions. 

Attorneys who intend to use an AI tool during a deposition or court proceeding must also evaluate whether the tool records audio or video, retains data, accesses a microphone or camera, or transmits confidential material. Real-time tools may not disrupt proceedings or be visible or audible to jurors. 

AI-generated or enhanced evidence 

The order establishes a formal disclosure process for substantive evidence generated or enhanced by AI. A party seeking to introduce such evidence ordinarily must give notice at least 90 days before trial and provide information about the technology, how it was used, the human review performed, its reliability, and the preservation of both the original and AI-processed evidence. A party challenging the evidence ordinarily must provide notice within 30 days after disclosure. 

AI-generated evidence remains subject to the ordinary rules governing relevance, authentication, reliability, and expert testimony. The court may also instruct the jury about the nature of the evidence and the process by which it was created. 

Experts and “black box” technology 

An expert who relies on AI must explain the tool, how it was applied, the nature of its algorithms and training inputs, and why the system is reliable. The order cautions that an expert relying on proprietary “black box” software may be unable to demonstrate reliability, reproducibility, error rates, or general acceptance. Opinions based on technology the expert cannot adequately explain are therefore likely to be excluded. 

Jury selection 

The order also addresses AI-assisted jury selection. A party intending to use AI for juror profiling, social-media analysis, or selection strategy must disclose that use before voir dire. Although attorneys may use certain programs to review publicly available information, they may not use AI-driven communications, targeted advertising, or microtargeting to reach or influence jurors or prospective jurors. 

The Court discloses its own use of AI 

In August 2026, Judge Smith issued an addendum addressing her own use of AI. The addendum allows the court to use AI as a productivity tool for outlining, editing, formatting, organizing written work, summarizing identified materials, and performing administrative tasks. It draws a firm distinction, however, between assistance and adjudication: weighing evidence, assessing credibility, making factual findings, interpreting the law, and deciding cases remain exclusively judicial functions. 

The court independently reviews all legal authorities, quotations, and citations before including them in an order or opinion. The same restrictions apply to law clerks, research attorneys, judicial assistants, and other personnel working under the judge’s supervision. 

Judge Smith explained that transparency about the court’s own practices was particularly important because she was asking lawyers to be similarly conscious of how they use AI. She described AI as a productivity tool that can assist with research, organization, and testing legal analysis, while emphasizing that it cannot establish facts or decide how the law applies to them. 

What businesses and their lawyers should do now 

As additional jurisdictions introduce policies of their own, lawyers and their clients must remain vigilant to ensure they are in compliance. Here’s how to approach this:

• Check the assigned judge’s standing orders early. AI requirements may vary not only by jurisdiction, but by courtroom. 
• Identify every place AI may touch a matter. Its use may arise in discovery, document review, medical-record summaries, demonstratives, expert analysis, depositions, and jury research, not just brief writing. 
• Understand how the tool handles data. Before entering client information, determine whether the platform stores prompts, transmits data externally, or uses information for model training. Independently confirm that the tool is closed source and does not breach confidentiality orders; don’t just rely on the AI vendor’s representations.
• Build human verification into the workflow. Lawyers remain responsible for facts, citations, legal conclusions, discovery responses, and evidentiary submissions regardless of how those materials were created. 
• Preserve original materials. When AI is used to alter, clarify, summarize, or enhance evidence, the original material and information about the process may be critical to admissibility. 

• Ask outside counsel about their AI practices. Clients should understand which tools their lawyers use, what information is entered, how outputs are verified, and whether a court order requires disclosure. 
• Build questions on AI use into your expert witness interviews and arrange the parameters of such usage of AI tools as part of the engagement process.

Judge Smith’s orders do not treat AI as inherently improper or incompatible with legal practice. Instead, they apply familiar principles of competence, confidentiality, candor, reliability, and professional accountability to new technology. The central message is straightforward: AI may assist the work, but it does not assume responsibility for it. 

Wondering how to build AI policies that anticipate potential rules? Reach out to Erin Mayer or another member of LP’s Litigation Group.