Can a Judge Use AI? The Ethics Lines Courts Are Drawing

When a lawyer uses AI, the professional responsibility framework is clear in at least one direction: the attorney bears personal responsibility for everything filed with the court. The lawyer’s name is on the pleading, and Rule 11 attaches. But what happens when it is the judge using AI? The judge is not filing pleadings — the judge is deciding cases. And that raises a different and more fundamental set of questions about the nature of judicial power, constitutional delegation, and the integrity of the adjudicative process.

Judicial AI ethics is one of the least-discussed but most consequential frontiers in the current AI-and-law moment. Courts are drawing lines, state ethics bodies are issuing guidance, and individual judges are making decisions about AI use in real time — often without formal guidance. What they decide affects every party who appears before them.

This post surveys the current state of judicial AI ethics guidance, the specific rules and standards that are emerging, and what the developing framework means for attorneys and businesses involved in litigation.

The Core Tension: Judicial Independence and Technological Assistance

Judges have always used tools. Law clerks draft opinions; legal research services provide case summaries; court staff manage dockets. None of this has been treated as a delegation of judicial authority, because the judge reviews the work, exercises judgment, and takes responsibility for the final product.

AI sits uncomfortably at the edge of this tradition. On one hand, using AI to research case law, organize document collections, or draft an initial structure for an opinion looks a lot like using a law clerk or a research service. On the other hand, a generative AI system does not merely retrieve and organize — it generates. It can produce analysis, apply legal frameworks, and reach tentative conclusions, all in ways that are difficult for the judge to fully audit. The risk is not that the judge is using a tool; it is that the tool is doing something that looks like deciding.

That distinction is where judicial AI ethics guidance is focusing.

Ohio’s 2026 AI Ethics Guide: Direct Rules for Judges

Ohio’s June 2026 AI Ethics Guide from the Ohio Board of Professional Conduct is unusual in that it addresses both lawyers and judicial officers in a single integrated document. The guidance for judges is the most specific state-level judicial AI ethics guidance currently available.

No AI for Ultimate Case Decisions

The guide’s most direct prohibition is on using AI to make ultimate case decisions. Under Ohio’s Judicial Conduct Rule 2.7, a judge must perform judicial duties “without bias or prejudice.” The guide interprets this in light of AI’s known failure modes — including hallucination, training data biases, and inconsistent performance across demographic groups — and concludes that the core adjudicative function, applying law to facts and reaching a reasoned conclusion that resolves the dispute, must remain entirely human.

This prohibition is absolute as a matter of principle. The guide acknowledges that AI can assist judges with research, document organization, and clerical tasks. But the moment the AI is generating the reasoning that resolves the case, the judge has crossed a line. The decision must be the judge’s decision, not a machine’s decision ratified by a signature.

The guide also cautions against using AI to produce the first draft of a judicial decision. This guidance is somewhat more nuanced — not a categorical prohibition — but it reflects a concern that AI-generated first drafts may shape the analysis in ways that are difficult for the reviewing judge to detect or correct. A judge who reviews and edits an AI-generated opinion may unconsciously accept the AI’s analytical framing even where independent judgment would have led to a different analysis. The guide’s recommendation is that AI assistance in opinion drafting, if used at all, be confined to research support and organizational assistance rather than substantive legal analysis.

The Ex Parte Information Problem

Ohio’s judicial AI guidance identifies ex parte communication risk as a distinct and serious concern. Ohio’s Judicial Conduct Rule 2.9 prohibits judges from initiating, permitting, or considering ex parte communications — communications about a pending case that occur outside the presence of the parties.

The ex parte risk from AI arises in a specific way. When a judge queries an AI tool about the facts or law of a pending case, the AI’s response may draw on sources, materials, or analytical frameworks that the parties have not had the opportunity to address. The AI may incorporate background information, related cases, or general context that was not presented in the briefing. That information — even if technically available in some public database — becomes problematic if the judge considers it in a manner the parties cannot contest.

The Ohio guide states that a judge should not use AI tools to independently conduct factual investigations that may result in the judge gaining personal knowledge of disputed facts. This mirrors the existing prohibition on judges conducting independent research into disputed factual questions, which is a well-established limitation on judicial authority. AI does not create a new exception to this rule; if anything, it makes compliance harder because AI-assisted research is faster and more comprehensive, creating a greater risk that judges will inadvertently acquire case-relevant information outside the record.

The guide also raises disclosure obligations: a judge who consults AI regarding a pending matter may be required to disclose that fact to the parties. This is an emerging area without settled authority, and the guide frames it as something judges should consider, but the direction of travel is clear — parties have a right to know if AI-generated analysis has influenced the judicial decision-making process, and to have an opportunity to respond to it.

Public AI Platforms: Confidentiality Concerns

The Ohio guide specifically cautions judges against using public AI platforms for case-related work. The concern is straightforward: a judge who enters case-specific facts, party information, or confidential details into a public AI platform may inadvertently disclose information that should remain within the proceedings. Many AI platforms retain user inputs, use them to improve the model, or share them under certain circumstances. A judge who queries “is a non-compete covenant enforceable in Ohio under these facts [detailed case facts]” using a consumer AI tool has potentially disclosed confidential case information to a third party.

This concern is particularly acute in cases involving sealed records, confidential protective orders, or sensitive personal information. The guide recommends that courts develop approved AI platforms with appropriate confidentiality protections for judicial use, rather than permitting judges to use whatever consumer AI tools they prefer.

Arizona: The First-in-Nation Judicial Technology Competence Rule

On August 27, 2025, the Arizona Supreme Court adopted a technology competence requirement for judicial officers — the first such rule in the country. Effective January 1, 2026, Comment 1 to Rule 2.5 of the Arizona Code of Judicial Conduct reads:

“Competence in the performance of judicial duties requires the legal knowledge, skill, thoroughness, and preparation reasonably necessary to perform a judge’s responsibilities of judicial office, including the use of, and knowledge of the benefits and risks associated with, technology relevant to service as a judicial officer.”

This rule change is significant precisely because it parallels the attorney competence rules that have long required lawyers to understand technology used in practice. Arizona is saying that judges have the same obligation: if you use AI, you must understand it well enough to use it responsibly.

The practical implications of Arizona’s rule are still being worked out, but the principle is clear. A judge cannot use an AI tool without understanding its hallucination risk, its training data limitations, or its known biases and then disclaim responsibility for any errors in judicial work product that AI contributed to. The technology competence requirement means judges are expected to evaluate AI tools with the same critical lens they apply to expert testimony — understanding the methodology, the known failure modes, and the limits of what the tool can reliably produce.

California: Mandatory Institutional AI Policies

California’s Judicial Council took a different approach. Rather than issuing rules for individual judges, the Judicial Council adopted Rule of Court 10.430, which requires California courts — as institutions — to adopt AI use policies. The rule became effective in 2025, with a compliance deadline of December 15, 2025, for individual courts.

Under Rule 10.430, courts that do not prohibit AI use by staff and judicial officers must adopt written policies governing that use. The policies must address AI use by court staff for any purpose and by judicial officers for tasks outside their adjudicative role. The California framework reflects a recognition that AI governance in courts is too complex to be addressed solely through individual judicial conduct rules — institutional policies are needed to coordinate training, platform selection, confidentiality protection, and oversight.

Los Angeles and Riverside County superior courts, among the largest trial courts in the country, adopted policies under this framework. Their current policies require disclosure only when a motion, decision, or other document is written entirely with generative AI. Partial AI use — using AI for research or initial organization but drafting the substantive analysis independently — does not require disclosure under current California court policies. This reflects a relatively permissive approach to AI-assisted judicial work that is at the narrower end of the emerging spectrum.

Federal Judicial AI Use: The Informal Landscape

Federal judges are not subject to state judicial conduct codes, but they are bound by the Code of Conduct for United States Judges. The Code has not been amended to expressly address AI, but its provisions on competence, impartiality, and proper use of judicial office provide a framework.

A 2026 random-sample survey of federal judges found that more than 60 percent of responding judges reported using at least one AI tool in judicial work, with legal research as the leading use case. Despite this widespread use, there is no formal federal judicial AI use policy. The Judicial Conference has issued informal guidance cautioning against delegating core judicial functions to AI and encouraging courts to develop local policies addressing AI use by court staff, but binding federal standards for judicial AI use remain absent.

The absence of formal rules does not mean no rules apply. Federal judges operate under the same constitutional limitations as all judges: they cannot delegate their adjudicative authority to a machine. The constitutional requirement that a judicial officer exercise independent judgment in deciding cases is implicit in Article III’s vesting of judicial power in courts staffed by judges with tenure and salary protections. An AI-decided federal court case would raise serious constitutional questions that no court has yet addressed, but that doctrine-watchers are beginning to discuss.

What the Framework Means for Practitioners

The emerging judicial AI ethics framework matters for litigators and their clients in at least four specific ways.

1. Disclosure and Objection Rights

When judges use AI in case-related work without disclosure, parties may be deprived of the opportunity to identify errors, challenge AI-generated factual findings, or address AI-introduced analytical frameworks. As judicial AI ethics develops, practitioners should be attentive to indications that AI has been used in a judicial decision — internal inconsistencies, unusual phrasing, factual references not in the record, or citations that do not match the proposition cited — and should consider whether disclosure obligations apply and whether objection or motion practice is appropriate.

2. Appellate Issues

Judicial AI use that violates applicable conduct standards — particularly undisclosed use of AI that incorporates information outside the record — may present appellate issues. These are novel arguments without established case law, but the principle that a judge’s consideration of extra-record information can constitute reversible error is well-established. AI use that effectively introduces extra-record information without party knowledge or consent may fit within that doctrine.

3. Recusal Considerations

In jurisdictions with disclosure requirements for judicial AI use, undisclosed AI use may raise impartiality concerns. If AI output has systematically influenced judicial reasoning in ways not disclosed to parties, recusal motions may be appropriate. Again, these are developing arguments without established precedent, but the doctrinal building blocks exist.

4. Advocacy Implications

Understanding that judges use AI for legal research means understanding that judges may first encounter your legal argument through an AI-generated summary rather than through your brief. If AI tools summarize your client’s position inaccurately, the judge may begin her analysis from a flawed premise. Practical implication: write briefs that are clear, structured, and AI-readable — with explicit thesis statements, clear headings, and careful use of defined terms — so that any AI-assisted review accurately captures your argument.

The Systemic Stakes

Beyond individual cases, judicial AI ethics raises systemic questions about the nature of the judicial process itself. Courts derive their legitimacy from a particular kind of authority: a human decision-maker, accountable under the law, who hears the parties and applies established legal principles through independent judgment. An AI system does not have that accountability, does not take an oath, cannot be impeached, and cannot be subjected to judicial discipline.

If AI decision-making infiltrates judicial processes without transparency — if parties appear before what they believe is a human judge but significant portions of the analysis are AI-generated — the legitimacy of those decisions is compromised in ways that are more than theoretical. Public trust in courts depends on the reality, and the perception, that human judgment is at work.

The guidelines emerging from Ohio, Arizona, California, and the federal judiciary reflect a collective recognition that AI in judicial work must be constrained, disclosed, and supervised. Getting those constraints right matters for every party who litigates in American courts.

Conclusion

Judicial AI ethics is one of the fastest-moving areas in the current AI-and-law landscape, and the rules are being written as the technology is being adopted. The current state of guidance reflects three core principles: judges may not delegate case decisions to AI; AI use creates ex parte and confidentiality risks that require careful management; and technology competence — understanding AI well enough to use it responsibly — is now part of what it means to be a competent judicial officer.

For practitioners, the practical implication is to understand the AI landscape in the courts where you appear, to be alert for indicators of undisclosed AI use in judicial work product, and to consider the advocacy implications of judges who may encounter legal arguments through AI-assisted research. As with so much in the current AI moment, the doctrine is developing faster than most practitioners realize.

This post is for general informational purposes only and does not constitute legal advice. Reading this post does not create an attorney-client relationship. If you have questions about your specific situation, consult a qualified attorney.