Alaap B. Shah, Member of the Firm in the Health Care & Life Sciences practice, in the firm’s Washington, DC, office, was quoted in the Washington Lawyer magazine, in “AI & Machine Learning Are Here. Will They Come for Lawyers?” by John Murph.
Following is an excerpt:
“The legal profession is undergoing a technological revolution, and ChatGPT is poised to play a major role in this transformation.”
So says ChatGPT when prompted to write a compelling argument on how it will disrupt the legal profession, touting its “ability to automate many routine tasks and provide quick and accurate answers to complex questions, freeing up valuable time and resources for lawyers and legal teams.” …
Unleashing Host of Possibilities — and Risks
In February 2023, a judge in Cartegena, Colombia, made headlines when he used ChatGPT, along with precedent from previous rulings, to decide whether an autistic child’s insurance provider should cover all of his medical treatment costs.
The same month, London-based law firm Allen & Overy LLP announced its partnership with OpenAI to launch Harvey AI, an adaption of ChatGPT to automate legal drafting and research. Reportedly, Harvey AI will work across many practice areas and languages to automate tasks such as contract analysis, due diligence work, litigation, and regulatory compliance.
There are several caveats that legal professionals need to consider before blindly relying on ChatGPT, including that its current version has only been tested with information up to 2021, with no access to any new laws passed since.
Alaap Shah, who specializes in privacy, cybersecurity, and data asset management matters at Epstein Becker & Green, P.C., asked ChatGPT whether abortion is a constitutional right. “ChatGPT got it entirely wrong because it only knows the world as it existed in 2021,” Shah says. “It does not have any awareness” about Dobbs v. Jackson Women’s Health Organization, which overturned Roe v. Wade in June 2022.
Another major red flag is the dubious citations that ChatGPT provides, as some lawyers have found. When Shah asked ChatGPT about a specific case, it gave him a citation that “looked very real” but that he could not find in either Westlaw or LexisNexis. “It was in the right nomenclature and everything,” Shah says. “I asked for another citation for another case holding a similar premise. Same thing. It gave me something that looked and felt real.”
According to Shah, no nonpublic sources were used in training ChatGPT, which means that it did not have access to traditional legal search tools such as Westlaw and LexisNexis. “One of the rubs of ChatGPT is there is no attribution,” Shah says. “It does not give you the location of where the information came from. At least for now, it doesn’t. You have no real way of verifying if the information is true or not unless you go to some secondary sources and do that diligence yourself.”
As if those inaccuracies aren’t troubling enough, lawyers who use ChatGPT also run the risk of violating client confidentiality. It’s wise to remember that ChatGPT is a learning-based AI system, which means that the more questions you ask, the more information it absorbs, legal tech experts point out.
