Like anyone else, judges need guidance in dealing with GenAI and LLMs. This Judicious Judge’s Guide is now available to do just that!

The Judicious Judge’s Guide to Generative Artificial Intelligence and Large Language Models was published last week and is available here.

The guide was co-authored by Texas District Judge Xavier Rodriguez, California Magistrate Judge Allison H. Goddard, University of Waterloo and Osgoode Hall Law School Professor Maura R. Grossman, Arizona Court of Appeal Division One Judge Samuel A. Thumma, and Louisiana Fifth Circuit (State) Court of Appeal Judge Scott U. Schlegel.

The paper is designed to provide guidance for judges and their chambers, using the ABA Model Code of Judicial Conduct (MCJC) as a foundation and help U.S. judges and their chambers use AI tools in ways that enhance, rather than erode, core judicial values.

The paper is broken down into six sections, with two appendices, as follows:

Sections I (Judicial Principles Governing GenAI Use) and II (Understanding GenAI and LLM Tools) provide basic technical and ethical guidance, discussing how GenAI should be treated as “a tool, not a decisionmaker; the judge must personally own all rulings issued in the judge’s name, a time-honored concept that long predates GenAI.”

Sections III (Ethically Permissible Judicial Uses of GenAI), IV (High‑Risk and Generally Inappropriate Uses of GenAI) and V (Governance, Safeguards, and Best Practices) offer guidance on how GenAI may be adopted responsibly within chambers:

• Particularly promising judicial uses include summarizing briefs, transcripts and exhibits; drafting routine orders and notices; editing judicial writing; creating timelines and issue lists; checking legal authorities; organizing documents; assisting with administrative tasks; and providing preliminary translation and transcription.
• High-risk and inappropriate uses include using AI to research parties, witnesses, or other facts outside the record; avoiding opaque algorithmic risk assessments from tools used in bail, sentencing or similar contexts; delegating substantive reasoning, such as making credibility determinations or weighing evidence; uploading sealed filings or sensitive information into consumer AI systems lacking appropriate confidentiality and data protections; and deploying AI agents without consultation with court IT and security professionals.
• Structured rules and practical guardrails to adopt GenAI responsibly include written policies for things like written guidelines that define permitted and prohibited AI uses, a list of approved tools, strict data-handling rules, and disclosure requirements; guidelines for vendor contracts, focus on tools with a “mosaic architecture” that constrains the model to generate from retrieved primary sources to aid in verification; and practical, checklist-driven protocols (including staff training) within their chambers to ensure consistent quality control and ethical compliance.

Section VI (GenAI-Related Issues Currently Arising in Litigation) discusses GenAI-related issues currently arising in litigation cases such as In re OpenAI, Inc., Copyright Infringement Litigation (where the Court ordered OpenAI to produce a sample of 20 million consumer ChatGPT output logs), U.S. v. Heppner, Morgan v. V2X, Inc., Warner v. Gilbarco, Inc., Jeffries v. Harcros Chemicals Inc., among others.

There’s also a discussion of AI-generated data as evidence: both acknowledged (including the status of Proposed Federal Rule of Evidence 707) and unacknowledged AI-generated evidence or AI-generated deepfakes. And, of course, a discussion about the rise of filings with AI hallucinations – from lawyers and pro se parties – and the rise of filings by those self-represented litigants. There’s even a discussion of prompt injections in filings, like this one we recently saw in the US and this one in Brazil).

As informative as the main sections of the Judicious Judge’s Guide are, its Appendices are just as useful for reference purposes:

Appendix A provides 19 specific judicial GenAI use cases, including summarizing briefs and transcripts, converting handwriting to text, drafting routine orders, proofreading opinions, checking citations, and much more. Each use case includes the typical tool type to use and key safeguards to apply when using them.

Appendix B collects existing federal, state and selected international judicial AI guidance, illustrating how courts are increasingly developing their own policies. While that list will be out of date quickly, it still provides several useful resources to consider regarding the use of GenAI by judges and participants in the court system.

The Judicious Judge’s Guide to Generative Artificial Intelligence and Large Language Models (available here) may be designed for judges, but it’s useful information for any of us participating in the legal system. Unless, of course, you don’t want to know what judges are thinking. 😉

So, what do you think? How well do you think judges are prepared for today’s GenAI-centric world? Please share any comments you might have or if you’d like to know more about a particular topic.

Image created using ChatGPT, using the term “robot judge reading a guide on generative AI”.

Disclaimer: The views represented herein are exclusively the views of the author, and do not necessarily represent the views held by my employer, my partners or my clients. eDiscovery Today is made available solely for educational purposes to provide general information about general eDiscovery principles and not to provide specific legal advice applicable to any particular circumstance. eDiscovery Today should not be used as a substitute for competent legal advice from a lawyer you have retained and who has agreed to represent you.