AI Legal Research for Busy Lawyers

Published in

walawlibrary

5 min read6 days ago

Legal news has been rife with discussion of new artificial intelligence (AI) products and predictions regarding the impact of AI on lawyers and the justice system. Large Language Model Generative AI products, whether generalist products, like ChatGPT, Microsoft’s Bing Copilot, or Google Gemini (formerly Bard), or law-specific variants like Casetext’s CoCounsel (now owned by Thomson Reuters after a $650 million acquisition), are designed to help users generate written text.

A graphic features the abbreviation “AI” in blue text, displayed centrally and prominently. The letters are surrounded on both sides by circuit board imagery, also in blue. A separate blue flowing graphic rises in the middle of the image over the letters. The image background is dark blue.

Lawyers have made headlines by asking generative AI to write complaints and trial briefs which are then filed without review and correction. One critical problem has been termed hallucination, which happens when the AI creates information, such as case citations, in order to make a document look like something that it has ingested. Generative AI is good at making documents that look like a known or commonly used format, but it is not a legal research tool and should not be used (or trusted) with legal analysis. At least not yet.

A corollary issue is the ethical implications of AI. Bar associations are beginning to release rules regarding the use of AI assistance. For example, the New York Bar Association recently advised that lawyers protect client confidentiality when they use AI. One approach is to add confidential client information after a first draft is generated by AI. This keeps the AI from ingesting client information. In November Mark Fucile wrote a good summary of these concerns in the Washington State Bar News entitled The Chatbot Made Me Do It! Confidence and Confidentiality in an Age of Artificial Intelligence. The Washington State Bar Association has appointed a Legal Technology Task Force to examine the impact of AI.

With those important caveats stated, AI holds great potential for busy lawyers. The Miami-Dade County Public Defender, an office of about 230 attorneys plus staff, has integrated CoCounsel into its workflow according to a December article in the Florida Bar News. Cindy Guerra, the Deputy Chief for Operations, explains:

Our lawyers are using it primarily for research, deposition preparation, first drafts. They’re using it to write memos. Everybody knows the story about the New York attorney, and the AI that made up cites. The attorneys know these are just first drafts that you still have to go through and make sure everything is right.

Unlike generalist large language models such as ChatGPT, CoCounsel ingests vetted law-specific information and was developed by a team of lawyers and computer scientists. Now that it is owned by Thomson Reuters, it also has, presumably, full access to the vast library of Thomson Reuters treatises and practice guides.

For lawyers who are not yet prepared to shoulder the added cost of CoCounsel, AI has been integrated into Westlaw Precision, Lexis+, and vLex/Fastcase (whose AI tool is called Vincent AI).

A laptop sits partially opened on a semi-reflective surface. The laptop monitor displays a bright multi-colored graphic and also displays a reflection of the keyboard. The image has been edited to bleed the different colors from the monitor in all directions to the edge of the image.

Lexis+ and Westlaw Precision Head to Head

The following is a comparison of the research results for AI assisted research for Westlaw Precision and Lexis+ for the question of whether officers need a search warrant to search a cell phone abandoned by a suspect.

Lexis+

Lexis+ has three AI tasks available to the user: Ask a Legal Question, Generate a Draft, and Summarize a Case. In 55 seconds the AI produced a three-paragraph answer noting the key U.S. Supreme Court case, Riley v. California, but without a citation (573 U.S. 373 (2014)), and the key Washington State Supreme Court case, State v. Samalia, 186 Wn.2d 262 (2016), that outlines the abandonment exception. It starts with the summary “In Washington State, law enforcement officers do not necessarily require a search warrant to search a cell phone that has been abandoned by a suspect.” It also cites State v. Garner, 26 Wn. App. 2d 654 (2023), State v. Fairley, 12 Wn. App. 2d 315 (2020), and State v. Priest, 2017 Wash. App. LEXIS 1805 (2017) (unpublished).

Westlaw Precision

Westlaw Precision has one AI task available to the user: AI-Assisted Research. In 53 seconds the AI produced about the same length of answer, citing the same cases as Lexis+. It also added cites to State v. McKee, 3 Wn. App. 2d 11 (2018), rev. State v. McKee, 193 Wn.2d 271 (2019) and two other unpublished Court of Appeals cases. Additionally, it cited a mishmash of trial court pleadings, a couple of law review articles about Riley v. California, and, more importantly, a citation to a section in Ferguson’s volume on Criminal Practice and Procedure in Washington Practice.

Both tools cited the key cases. Both tools gave a correct summary of the abandonment exception. Both tools warned of the privacy implications involved. Both tools saved time compared to keyword searching a Washington caselaw database and sifting through the results. Both tools keep getting faster and faster. However, reliability overall has been shown to be lackluster and neither AI is hallucination free.

In actual practice it would be important to use a citator (Shepard’s or Keycite) to double-check the cases that the AI identified, to see if there was negative history or newer cases on point. Researchers would also need to read the cases to make sure the AI was accurately summarizing their holdings. The devil is in the details, especially in regard to analogizing and distinguishing facts in relationship to holdings. Secondary sources, like Washington Practice, topical LexisNexis practice guides, or the WSBA Deskbooks, are useful in understanding legal doctrine and, unlike AI generated content, they are written by experienced local attorneys. AI legal research can be a helpful starting point but shouldn’t be the final step.

National Center for State Courts webinars and videos on AI are available here