When Generative AI Fabricates Cases That Attorneys Cite, Sanctions Follow | American Enterprise Institute
On June 22, 2023 a federal judge in New York City sanctioned two attorneys and their law firm for “abandon[ing] their responsibilities when they submitted non-existent judicial opinions with fake quotes and citations created by the artificial intelligence tool ChatGPT, then continued to stand by the fake opinions after judicial orders called their existence into question.” Judge P. Kevin Castel imposed a $5,000 total penalty on attorneys Peter LoDuca, Steven A. Schwartz, and their firm, Levidow, Levidow & Oberman P.C., in Mata v. Avianca. Castel concluded that LoDuca and Schwartz acted in “bad faith” when they did not immediately come clean about their reliance on ChatGPT and what Castel bluntly called “bogus cases.”
The sanctions, however, go beyond monetary punishment to include a dose of court-mandated outing of the attorneys’ actions to certain concerned individuals. Specifically, LoDuca, Schwartz and their firm must mail to their client (Roberto Mata) a copy of Castel’s opinion and order, which details their conduct, along with a transcript of a hearing regarding their use of AI. They must do the same for “each judge falsely identified as the author of the fake . . . opinions.”
Via Twenty20
The case drew national headlines in May when the story broke about the fabricated cases generated by ChatGPT. It illustrates at least three vital lessons: 1) generative AI technology, like any technology, is fallible and currently doesn’t function like a database search engine; 2) humans who rely on AI thus bear the ultimate responsibility for ferreting out its inevitable errors; and 3) the actions of a few may harm many.
Regarding the first lesson, the euphemism “hallucination” is gaining media traction when describing the propensity of generative AI to sometimes produce false information that masquerades as true. As an article in the Washington Post wryly put it, chatbots sometimes “act like precocious people-pleasers, making up answers instead of admitting they simply don’t know.” Attorneys (and anyone) using generative AI to assist in official business need to understand that. Ronald Minkoff, an attorney for Levidow, Levidow & Oberman, stated his firm’s attorneys didn’t know that, however, and simply “made a good faith mistake in failing to believe that a piece of technology could be making up cases out of whole cloth.”
But attorneys also must understand something else––ChatGPT currently does not function like a database search engine. It is a large language model. As Jeff Hermes, an attorney and deputy director of the Media Law Resource Center, explained last month when the story about Mata v. Avianca broke:
large language models are not search engines. They do not scour a database of existing resources for those that match our query. They are indeed based on massive amounts of information drawn from online resources, as are search engines, but what they draw from those resources is different. Specifically, they have used those sources to gather probability data regarding the sequencing of words and characters in order to produce an output that seems to be related to user inputs. They look at syntax, not semantics.
Regarding bottom-line human responsibility, it helps to recall that good senior partners in charge of a case––the lead attorneys––should always review the work of junior associates they assign to it. That same sense of personal, buck-stops-here accountability must apply whenever any attorney farms out work to a generative AI system.
As Mike Masnick wrote in TechDirt in early June about the actions of the attorneys representing Roberto Mata, “You can blame ‘generative AI’ all you want, but the real mistake here, for which there is no excuse, was their failure to actually check on the cases they referenced.” For Masnick, it amounted to “horrifically bad lawyering,” even if the attorneys were not malicious in their actions. Masnick pounded this point home in a later post after Judge Castel’s sanctions order, writing that “the real ‘error’ here was not so much the use of ChatGPT, but rather not doing the absolute most basic work of checking citations to see if they were still good law (which . . . should have turned up the fact that those cases didn’t even exist).”
Finally, as for the third lesson, Judge Castel was clear that the potential harm caused by submitting fake opinions stretches beyond the confines of the case in which they are tendered. More broadly, “it promotes cynicism about the legal profession and the American judicial system.” He added that “there is potential harm to the reputation of judges and courts whose names are falsely invoked as authors of the bogus opinions and to the reputation of a party attributed with fictional conduct.”
In sum, the case of the fabricated cases provides both an early cautionary tale for attorneys and a teachable moment about generative AI. Like AI itself, we can learn from our mistakes.
Source link
Attorneys and law firms can elevate their online presence with professionally written content from SEO Content Writing Services monthly plan. Our team of experienced writers specializes in crafting blog posts, articles, and written content that accurately reflect the expertise and knowledge of our clients in the legal field. With a keen eye for detail and a thorough understanding of legal terminology, we provide high-quality writing that helps our clients stand out from their competition and engage with their target audience. Trust SEO Content Writing Services for one time article writing or monthly written content to handle all of your written content needs and showcase your law firm's expertise.
If you need written content, blog posts, or articles professionally written for your website, we can help. Go HERE to find out more.
or email us here: myseowritingservices@gmail.com
To find out more about our article writing or blog post services, fill out the form, thank you.