by James Arden

James Arden has been involved in legal malpractice litigation and appeals since 1987. He is a member of LACBA’s Professional Responsibility and Ethics Committee, and the Association of Professional Responsibility Lawyers (APRL). He writes for other lawyers, and authors TechnoEthics for the ABA’s GPSolo magazine.

Articles are provided regularly by LACBA’s longstanding Professional Responsibility and Ethics Committee. 

“I heard about this new site which I assumed -- I falsely assumed was like a super search engine called ChatGPT, and that’s what I used.” That’s what attorney Steven A. Schwartz said to United States District Judge P. Kevin Castel as he tried to avoid being sanctioned for having filed papers citing more than half-a-dozen cases that did not exist.[1]

Schwartz had wanted to oppose a dismissal motion by showing Roberto Mata’s case wasn’t time-barred under the Montreal Convention. Schwartz asked Chat GPT to “provide case law,” “show me specific holdings,” and “show me more cases.” ChatGPT complied by making up cases.[2] Schwartz had made some further inquires of ChatGPT, but neither he, nor a second attorney in the firm, did anything to ensure the cases they cited really existed before they filed the brief.[3]

The two attorneys who signed the brief were hoist on their own petard because they used ChatGPT’s output in a brief for a client without understanding what ChatGPT does. Lots of lawyers and people are trying ChatGPT. Its website generated over 1.6 billion visits just in June 2023.[4] But attorneys, having express duties of competence and confidentiality, have to understand how ChatGPT processes what is input before typing in any client information.

When one first lands at ChatGPT’s website, a click-through screen warns that whatever is input may be reviewed by other people; and advises not to share sensitive information. A second screen warns: “the system may occasionally generate incorrect or misleading information” and “It is not intended to give advice.”

The attorneys who were sanctioned in the Mata case filed a brief without having read the cases they cited.[5] No attorney should cite a case she or he hasn’t actually read, even if the case came from a learned treatise or annotated code. But the real problem these attorneys had was that they had no idea what they were doing with ChatGPT: “I just was not thinking that the case could be fabricated, so I was not looking at it from that point of view.”[6]

Because of their misapprehension, they cited a number of fabricated cases. “Counsel should never misrepresent the holding of an appellate decision. Not only would that be a violation of counsel’s duty to the court (Bus. Prof. Code, § 6068, subd. (d)), it will backfire because the court will discover the misrepresentation, particularly when it relates to a decision issued by that court.”[7]

The only thing these attorneys seemed to know about ChatGPT was that it was a website where they could try using AI (artificial intelligence.) ChatGPT is just one source for one type of AI. It is a chatbot, a bot that acts as though it chats with you. You’ve encountered chatbots on many commercial websites. A small screen will pop up, and “Emily” or “Larry” or “Topper” will cheerily ask if it may assist you.

So why is everyone talking about ChatGPT? There are many other chatbots. The answer: ChatGPT, from a company called OpenAI, is free. You just go to ChatGPT.com, and set up a free account.

There are also chatbots tailored to the legal profession. All the publishers catering to lawyers have them available through their websites, but they’re not free. That’s why so many lawyers try ChatGPT first. Most lawyers though, unlike the two mentioned above, don’t start using new technology without understanding its rudiments.

As Bill Gates explained: “current AI models …. aren’t necessarily good at understanding the context for a human's request, which leads to some strange results. When you ask an AI to make up something fictional, it can do that well. But when you ask for advice about a trip you want to take, it may suggest hotels that don't exist. This is because the AI doesn't understand the context for your request well enough to know whether it should invent fake hotels or only tell you about real ones that have rooms available.”[8]

Chatbots designed for lawyers draw from up-to-date legal databases for answers to queries. Lawyers using ChatGPT need to know a little about how it works and understand that the sources from which it derives its responses are limited.

ChatGPT does not search the Internet like Google or other search engines to find answers. It does not access a database of facts. Instead, its responses are based on patterns it saw in its training data. ChatGPT was trained on a variety of books, news articles, scientific journals, even Wikipedia. But its knowledge has been limited to dates, events, and people prior to about September 2021.

Just this year, ChatGPT began experimenting with an actual web-browsing plugin that will allow it to draw on information from around the web.[9] Even so, it won’t be the same as an Internet search engine.

Here’s why. Santiago Paz of Dentons writes: “Chat GPT will always return an answer even if there is no answer to the prong asked by the user within the ChatGPT universe of information on which it was trained for. This means that if there is no answer to the question asked by the user, it will return a wrong answer.”[10]

Another limitation: ChatGPT is incapable of resolving nuanced situations. It does not reason like a human being. It returns answers based on statistical patterns of data on which it was trained. Asking it about situations that are new and unique, or without precedent “will most likely lead to the opposite or wrong answer.”[11]

Also, ChatGPT has no capacity to interpret or analyze legal documents or complex legal provisions, like an experienced lawyer.[12]

Lawyers using ChatGPT have to be concerned about its potential ethical risks too. The most important one is confidentiality. Suppose you want to create The Best Brief Ever. If what you input to ChatGPT refers to a confidential matter or communication, you might have just waived attorney-client privilege and breached Rule 1.6 if you did not have consent from your client to reveal the matter.

Or you want to revise a contract. You copy the text of a draft into ChatGPT. It finds several issues, you revise the contract, and ChatGPT produces a New and Improved version of it. Lance Eliot of Forbes notes, “Behind the scenes and underneath the hood, the contract might have been swallowed up like a fish into the mouth of a whale. … Your prompt as provided to the AI app is now ostensibly a part of the collective in one fashion or another. … The outputs of ChatGPT are also a type of content that can be retained or otherwise transformed by the AI app.”[13]

Paz of Dentons points to even another type of concern for ChatGPT’s security. During a Chat GPT security breach earlier this year, users could see other users conversations.[14]

Other than privilege and confidentiality, and competence (e.g. the lawyers above,) ChatGPT poses potential issues as to billing ethics. Since the advent of word processing, lawyers have able to reuse work product–but they may not bill for recreating wheels. “A lawyer who is able to reuse old work product has not re-earned the hours previously billed and compensated when the work product was first generated.” (ABA Form, Op. 93-379, Dec. 6, 1993.) So, a lawyer who spends a few minutes using ChatGPT to generate a 15-page Best Brief Ever, and who then spends two hours rewriting and editing it–and checking citations, shouldn’t charge the client more than two hours-and-a-few-minutes, no matter that it might have taken ten hours had The Brief been drafted by the lawyer personally.

I thought I’d ask ChatGPT, what did it “think” was the best way to research the law? It responded, literally, with “a step-by-step guide on how to research the law.” It suggested, among other things: identify your research goal; start with secondary sources; use online legal databases; search for primary sources; analyze case law; review statutes and regulations; check validity and currency; cite sources properly; and cross-check information. If only Roberto Mata’s lawyers had done some of that.

“Technology makes it possible for people to gain control over everything, except over technology.” John Tudor.