This summer I followed the stories about the New York lawyers who got sanctioned for submitting ChatGPT-generated briefs with cites to non-existent cases with a certain amount of schadenfreude. But I was also, secretly, curious – how does one generate an entire brief from ChatGPT and what would it look like? So, naturally, I tried to harness the new AI technology for myself. My results were not quite what I expected – I came nowhere close to getting any sort of legal analysis or even substantive case law out of ChatGPT (how did those New York attorneys even do it?), but successfully used ChatGPT revise and improve my own, already-written work product.
First, I asked ChatGPT for caselaw on economic duress (having recently briefed an issue involving economic duress, I was familiar enough with the defense to be able to spot-check ChatGPT’s answers). ChatGPT apologized and explained that it doesn’t “have access to real-time legal databases or the ability to provide specific case law information beyond my last training data in September 2021.” I asked whether I could even use ChatGPT to find legal citations to cases, and ChatGPT assured me that I could. It recommended “requesting a list of relevant cases related to a certain area of law.”
Accordingly, I asked for “a list of relevant cases related to the economic duress doctrine.” But again, ChatGPT demurred, telling me that it doesn’t have “real time access to current legal databases of the ability to provide specific lists of cases.”
I changed tactics and asked ChatGPT if it could help me generate a brief explaining why the economic duress doctrine is not available when the threat was filing a lawsuit. This time, ChatGPT gave me an outline of a legal brief to “get me started,” provided the caveat that it was a “general outline” only, and recommended I consult with an attorney. While the outline correctly stated the general law regarding economic duress, including why a lawsuit is not a sufficient threat to create duress, it was a far cry from a brief I could submit to court.
At this point, I abandoned my attempts to have ChatGPT write me a legal brief. I was both relieved that generating a legal brief with ChatGPT was not exactly easy (again, how did those New York attorneys manage it in the first place?) and also a little annoyed that I lacked the skill to convince ChatGPT to give me legal cites, imaginary or real (did I not ask politely enough?).
I found significantly more success when I asked ChatGPT for help with smaller, more discrete writing tasks. ChatGPT was quite helpful with refining paragraphs and clunky sentences. For example, when I asked ChatGPT to help me concisely refine a paragraph I had previously written as a summary of the changes to Colorado’s new non-complete law, I was pleasantly surprised with the results.
My original paragraph read:
All told, while the new law explicitly affirms that the general assembly intends to “preserve existing state and federal case law in effect before the effective date of this act,” there is no denying that the landscape has shifted. Going forward, in addition to ensuring compliance with the notice requirements and compensation limitations, employers should carefully consider the scope of the restriction. Further, these changes may encourage employers to consider other means of limiting unfair competition in certain circumstances.
ChatGPT suggested the following, better, paragraph:
While the new law confirms the intent to preserve existing state and federal case law before the effective date, it’s evident that the landscape has changed. Moving forward, employers must ensure compliance with notice requirements and compensation limits, while also thoughtfully assessing the scope of restrictions. These changes may prompt employers to explore alternative ways of curbing unfair competition when appropriate.
ChatGPT was also useful for coming up with helpful synonyms or even analogies. When I asked ChatGPT to help me come up with a synonym for the phrase “red herring” for use in a motion, ChatGPT suggested using, among others, “distraction,” “smokescreen” and “misdirection,” all of which would be appropriate in a motion depending on the tone you are seeking.
When tasked with coming up with an analogy for the idea that while you can present alternative legal theories, you cannot present alternative facts, ChatGPT suggested the following:
“Presenting alternative legal theories in a case is like offering different paths to reach a destination using a map. Just as travelers can choose different routes to the same place, attorneys can advocate for distinct legal perspectives within the framework of the law. However, asserting alternative facts is akin to suggesting that there are multiple versions of the landscape itself. The landscape remains constant, and while interpretations of it may vary, the fundamental facts remain fixed and immutable.”
I decided to ask ChatGPT how I should conclude this article when my conclusion was that ChatGPT did not appear to be capable of generating entire legal briefs out of whole cloth but was quite useful for discrete writing and editing tasks. In response, ChatGPT gave me the following suggestion for a conclusion:
“In conclusion, our exploration of ChatGPT’s capabilities in the realm of legal writing has shed light on its strengths and limitations. While ChatGPT may not yet possess the capacity to craft entire legal briefs from scratch, its value shines through in the realm of discrete writing tasks and precision editing. As technology continues to evolve, the integration of AI tools like ChatGPT into legal workflows proves invaluable for enhancing efficiency and accuracy. However, it is crucial for legal professionals to leverage AI as a complementary tool, harnessing its strengths while maintaining their central role in providing legal expertise and nuanced analysis. The synergy between human legal acumen and AI-driven assistance holds the promise of a more dynamic and effective future for the legal profession.”
I couldn’t have said it better myself.