A Review of Sedona’s “Artificial Intelligence (AI) and the Practice of Law” by The Hon. Xavier Rodriguez

Additional Monetary Sanctions in Freeman v. Giuliani
September 26, 2023
Bob Dylan Awarded Discovery Sanctions After Dismissal of Tort Claims
October 4, 2023

“Artificial Intelligence (AI) and the Practice of Law,” by the Hon. Xavier Rodriguez, is available at 24 Sedona Conf. J. 783 (forthcoming 2023).[1]  The article focuses on issues that “practicing attorneys are likely to encounter and steps state bars and related entities should consider.” Id. at 786.  Sedona’s website states:

Judge Rodriguez presents a comprehensive, broad-based examination of how the rapid growth of artificial intelligence is poised to reshape the legal profession. AI has potential to reduce the costs and improve the quality and availability of legal services by automating routine functions and allowing lawyers to spend more time on case analysis and crafting legal arguments, but it will require attorneys, courts, rules committees and ethics bodies to assess the risks and challenges associated with its use.

The article seeks to provide attorneys with a baseline understanding of AI technology and recommends areas where state bars, courts, rules committees, and attorneys may wish to undertake further study and potential rules changes. It can be downloaded from the website….

Judge Rodriguez’s excellent article begins with the statement that “the adaptable lawyer will need to determine when and if to incorporate AI into his or her practice….  AI is poised to re-shape the legal profession. But AI will require courts, rules committees, and ethics bodies to consider some of the unique challenges that AI presents. It will require attorneys to evaluate whether to use such products, and the risks associated with any use.”

The Judge begins the Sedona article with an introduction to AI, explains that it is already “ubiquitous,” and suggests that attorneys need to have a basic understanding of it.  It then provides a definition and clear and concise roadmap to that understanding.  E.g., id. at 788, passim.  Further, the article describes how major legal research and e-discovery vendors, as well as law firms, are incorporating it.  Id. at 789-91.

Noting the many advantages and uses of AI, the Sedona paper also explains potential limitations on current generative AI programs. 24 Sedona Conf. J. at 791, passim. For example, and without limitation: “Was the data used to train the system skewed or incomplete?…  Most importantly, was the AI system specifically designed to be used by lawyers and the legal profession?”  Id.  It points to both risks and advantages in using AI to prepare motions and briefs. Id. at 799, passim.

What remains unclear is whether AI platforms are equivalent to a nonlawyer requiring supervision, as contemplated by ABA Model Rule of Professional Conduct 5.3. It is also uncertain whether negligent reliance on AI tools can establish a violation of these rules, and whether lawyers must exercise “supervisory authority” over the AI platform, such that the lawyer must make “reasonable efforts” to ensure that the AI platform’s out-put is compatible with the attorney’s professional obligations.

Id. at 800.

An entire section of the Sedona article addresses AI and eDiscovery. Id. at 810, passim.  It states: “How generative AI and LLMs will be incorporated into eDiscovery remains uncertain.”  The Judge wrote that:

It is expected that the natural language search capabilities of LLMs will be incorporated into eDiscovery platforms at some point. This will allow AI to recognize patterns and identify relevant documents. Unstructured data (e.g., social media and collaborative platforms like Slack or Teams) can be reviewed by the AI tool. Theoretically, collection and review costs could be dramatically lessened, and attorney fees reduced. Another possibility is that AI will be used to augment the document gathering and review process, as well as assist with privilege review. For example, the Clearbrief platform, among others, is already being used for this purpose, with the underlying source documents visible in Microsoft Word so the user can verify the AI suggestions of documents. Users can then share a hyperlinked version of their analysis with the cited sources visible so the recipient can also verify the relevance of the source document.

Id. at 811.

The Sedona article addresses the need to protect client confidentiality.  24 Sedona Conf. J. at 796, passim. For example: “Attorneys considering using AI platforms should take care not to disclose confidential information inadvertently by inputting such information into a prompt or uploading confidential information into the AI platform for processing, particularly when the AI system is open source, such as the free version of ChatGPT, and the terms of service may not guarantee confidentiality.”  Query history may be saved by the AI system, and that could present ethical issues. Id. at 797.

Judge Rodriguez proceeds to address the many potential opportunities that AI offers. Id. at 795, passim.  He writes that:

AI tools offer the prospect to automate and possibly improve several operations, including legal research, document review, and client communication. The use of AI could also free lawyers to work on issues of strategic importance—both improving the experience of practicing law while at the same time providing more value to the client. In addition, AI’s ability to analyze large amounts of data can reduce the risk of human error and increase confidence in the accuracy of the results produced.

Addressing the problem of “hallucinations,” the Sedona article notes that models developed using archives of legal documents may improve accuracy. Id. at 795.  At bottom, the article notes that AI has the potential of decreasing legal costs; however, it will not replace human verification and “human judgment.” Id. at 796.  The Judge wrote that:

It is expected that AI tools will be able to: (1) facilitate alternative dispute resolution (ADR) by providing early insights into disputes, (2) predict case outcomes, (3) engage in scenario planning and predict negative outcomes, (4) assist with case management and calendaring/deadlines, (5) conduct contract review and due diligence tasks, (6) automate the creation of forms and other legal documents, (7) assist with discovery review and production, (8) assist with the ability to detect personal identifying information, confidential health information, or proprietary or trade secret information, (9) enhance marketing and social media presence, (10) translate data into another language, (11) automate billing, and (12) expedite and lower the cost of legal research and regulatory compliance. In addition, counsel may be able to use AI tools to engage in strategic planning with their clients by running analyses of the client’s financial statements and other data. That said, many other non-AI tools can assist with these tasks. Ultimately, attorneys and clients will need to evaluate whether the benefits of this new technology outweigh any costs or privacy or security concerns.

Id.  The Judge also notes that: “AI-powered redaction tools could possibly automatically identify personally identifiable information (PII) and redact that material from large data sets. AI-powered redaction tools reduce the risk of accidentally disclosing sensitive data because of human error.”  Id. at 798.

For an example of the use of AI in e-discovery, Kelly Twigger and eDiscovery Assistant recently announced the roll out of generative AI summaries into their case law database.  Kelly wrote: “After months of prompt engineering to focus in on the specific ESI issues in a decision, all 32,000+ decisions in our case law database will display a summary for our users in both search results and on the case page to allow them to assess the applicability of a decision quickly. First we reduced finding the relevant cases to a few clicks, and now we’ve made them faster to analyze.” eDiscovery Assistant Unveils AI-Generated Case Law Summaries with Cutting-Edge Generative AI Technology | eDiscovery Assistant

One risk mitigation strategy for lawyers and law firms not using a database such as eDiscovery Assistant is implementation of AI policies.  Id. at 798, passim.  This may include anything from guidance to a total ban.  Id.  However, Judge Rodriguez wrote:

A better approach may be to instruct employees that they are responsible for checking any AI’s output for accuracy, they should consider whether the output of any AI platform is biased, that all appropriate laws be complied with, and they evaluate the security of any AI platforms used before inputting any confidential information.

Id. at 798-99.

The comprehensive Sedona article addresses evidentiary issues connected with AI. Id. at 801, passim.  For example, AI evidence may be challenged under authenticity and other traditional evidentiary concepts.  Id.  “AI evidence may require that the offering party disclose any training data used by the AI platform to generate the exhibit.”  Id. at 802.  Judge Rodriguez cites a split of authority on authentication when proprietary algorithms are involved. Id. The cited cases involve use of AI in sentencing in a criminal case and use of AI in an appraisal system for teachers.  Id. He concludes that: “AI evidence requires a balancing between protecting the secrecy of proprietary algorithms developed by private commercial enterprises and due process protections against substantively unfair or mistaken deprivations of life, liberty, or property.”  Id.[2]

Judge Rodriguez wrote that: “To adequately counsel clients, attorneys will need to keep abreast of regulatory and statutory developments in this area.”  Id. at 812.  His Honor describes the rapidly changing regulatory context, as well as the impact on individual privacy.  Id. at 812, passim.

Perhaps the most interesting and important section of the paper is “Additional Training or Skillsets Required.” Id. at 817, passim.

If AI tools are used, AI should be used to complement human judgment. Lawyers and legal professionals should contemplate how to leverage this collaboration effectively and efficiently.  Prior to using any AI tool, lawyers should consider what processes currently used could be improved through AI technology. If AI tools are adopted, personnel will likely require training on how to properly construct prompts/queries and how to evaluate any results. Akin to Boolean searches that require some knowledge of how to construct a “good” search, AI tools require “good” prompts. One advantage of generative Ai prompts and responses is that the tool has “thread” conversations. A person can ask clarifying questions. Users can ask the AI tool to clarify previous responses or ask the AI tool to customize the tone or persona of the response. Personnel will also need training on compliance with confidentiality concerns, as well as considerations involving bias. Some commentators envision a new category of employee being trained–a “prompt engineer.”

Id. at 817-18.  In addition to the potential ethical and billing considerations addressed, id. at 820, the paper discusses minimum continuing legal education, id. at 820,  the need for law school classes, id. at 821, passim, and the need for judicial training, id. at 822, passim.

It has always bothered me that some courts have entered orders requiring disclosure of the use of AI, and the Sedona paper states:

[S]ome judges have required the disclosure of any AI that the attorney has used. As noted above, that is very problematic considering how ubiquitous AI tools have become. Likely these judges meant to address whether any generative AI tool had been used in preparing a motion or brief. That said, if any order or directive is given by a court, it should merely state that attorneys are responsible for the accuracy of their filings. Otherwise, judges may inadvertently be requiring lawyers to disclose that they used a Westlaw or Lexis platform, Grammarly for editing, or an AI translation tool.

Id. at 822.

At bottom, the Sedona paper provides excellent, readable, and comprehensive coverage of a fascinating and emerging issue.  Judge Rodriguez notes that:

Courts probably will not wish to pursue a “smart court” model of justice now being implemented in some Chinese cities. In the latter model, AI tools generate pleadings for litigants, analyze the litigation risk and issue a judgment—all done virtually.

Id. at 824.   The Sedona article set out to provide a baseline roadmap of the advantages and disadvantages, as well as the proper uses and risks presented by this marvelous new technology.  It has done a superb job of that.  All of The Sedona Conference’s great publications can be downloaded for free.  Publications | The Sedona Conference®

In a review of another book, I recently quoted the old military maxim that Any Ship Can Be a Minesweeper – – –  Once.  In that blog, I quoted Ms. Laura Zubulake’s comment that, where there is e-smoke, there is good reason to suspect an e-fire.[3]  Judge Rodriguez’s article shows how not to step on the mine or get burned by the fire.


[1] Judge Rodriguez extended special thanks to a number of other contributors who are listed in the publication.

[2] The article addresses many other issues that are not included in this review, such as, for example, “deep fakes.”  Further, because this blog focuses on civil discovery, the discussion of AI in law enforcement and the criminal justice system is not covered here.  See, id. at 806, passim.  Further, the Sedona paper addresses specialized practice areas, such as AI and employment law,  id. at 809, passim, health care law, id. at 811, passim, immigration law, id. at 812, alternative dispute resolution, id. at 815, passim,  and law firm marketing, id. at 817, passim.  The article points out that AI may assist non-attorneys and pro bono attorneys, and help close the access-to-justice gap.  Id. at 814, passim.  AI and cybersecurity is discussed. Id. at 818, passim.

[3] L. Zubulake, Zubulake’s e-Discovery (2014), 44.