At the request of his father, Austin attorney Dov Preminger took on a case a few years ago for a contractor who didn’t get paid about $50,000 for work he performed.
Preminger’s work on the case came to about $13,000 in fees and expenses. While he and his client won a default judgment for the full requested amount, they were only able to garnish a bank account for about $3,000, which was split between the client, the bank’s lawyer and Preminger.
Preminger tells that story to illustrate the potential usefulness artificial intelligence can provide to lawyers and clients as he gradually launches LitLaw, a “legal AI company” that instantly generates first drafts of lawsuits, answers, demand letters, discovery requests and more. LitLaw will eventually be able to draft more complex motions, Preminger said.
“That’s how litigation is a lot of times: It’s really, really expensive and inefficient,” Preminger said. “So the idea is to try to alleviate some of that and make it cheaper and make it affordable to everyday people.”
Preminger touted LitLaw as a first of its kind. Unlike research tools like Thomson Reuters’ Westlaw and LexisNexis’ Casetext which create memos, LitLaw will draft a pleading and return it in Microsoft Word, ready for a lawyer’s edit. Thomson Reuters this month announced a similar function called CoCounsel Drafting that produces transactional documents in Microsoft Word, with litigation documents to come later this year.
LitLaw will be a subscription service; the team hasn’t settled on prices yet, but Preminger said the cost will likely be a few hundred dollars monthly.
Preminger and his engineering partners, Abraham Adberstein and Jacob Adberstein of Phantom Labs, are using Claude, a large language model like its more commonly known competitor ChatGPT, but that Preminger thinks is the best writer. LitLaw pulls its knowledge from what’s publicly available on the Internet, they said, and their team has fed into it Preminger’s petitions, discovery requests and other pleadings to finetune it.
The quality of the drafts mirrors what a mid-level associate might produce, said Preminger, who opened his firm in 2022 after working for about six years at Norton Rose Fulbright.
Preminger demonstrated how the technology works in an interview with The Texas Lawbook. He fed information of a fictitious case in which one of his tech partners sued the other for breach of contract to create a demand letter and a lawsuit. In less than a minute, the draft petition included a usual preamble, discovery, parties, jury demand, jurisdiction and factual background. The technology chose breach of contract and fraud as its causes of action. Preminger said he would probably tell the program to add promissory estoppel to the petition.
“But this is a pretty ready-to-go complaint as far as I’m concerned,” Preminger said. “This is probably something that would take an associate at least a couple of hours to do. The idea behind this is now you just get that draft instantly.”
Preminger’s friend and lawyer, Will Langley, was inquisitive before signing up for LitLaw. Concerns of the technology hallucinating or sharing confidential client information were front of mind.
Langley, who was an associate at Winston & Strawn before launching his Houston-area consulting firm last year, is no stranger to AI tools. He worked with a friend to create a downloadable interface so sensitive information is not stored online.
“I think you’d have to be bold to say you’re not nervous about [hallucinations] at all,” Langley said.
Lawyers have retold horror stories of two New York lawyers who faced sanctions for submitting AI-generated legal briefs that included fictitious cases. A Stanford University study of LexisNexis’ and Thomson Reuters’ legal AI tools concluded in May that each hallucinates between 17 and 33 percent of the time.
These so-called hallucinations led U.S. Supreme Court Chief Justice John G. Roberts Jr. to caution the legal community in his most recent year-end report. In Texas, at least two judges have standing orders requiring lawyers to vet the work of AI if it is used: U.S. District Court Judge Brantley Starr in Dallas and State District Court Judge Roy Ferguson whose jurisdiction covers five southwest Texas counties.
Milan Markovic, a Texas A&M University School of Law professor, is skeptical of the hype for AI tools in the legal industry. Markovic, who concluded AI is unlikely to replace lawyers in his piece “Rise of the Robot Lawyers?” in the Arizona Law Review, said the tools may be good for rote tasks like summarizing documents and creating demand letters. He noted, however, that lawyers have used templates for basic pleadings since long before the creation of AI to question how much time the AI will actually save.
Markovic also emphasized concerns about sharing client information with AI tools and questioned how much is retained to train the technologies.
“AI certainly does have some really good utilities. My concern is, I think we’re overstating them,” Markovic said. “The ethical obligations of attorneys to ensure that they don’t have things like hallucinations and everything else are going to undercut a lot of these supposed efficiencies.”
To Langley, the friend of Preminger’s who intends to use LitLaw, the technology stands to save him valuable time. He contends the time spent fact-checking AI-produced drafts is still less time-consuming than writing from scratch.
As for data privacy, Preminger was able to assuage Langley’s concerns.
Preminger said his team has a zero data retention agreement with Claude, which dictates Claude does not train its AI on LitLaw’s data and Claude does not retain any of the data after it has processed the request.
“You can be confident that your data is safe,” Preminger said.
LitLaw’s beta version does not create complex motions, so it does not draw on case law and hallucinations are not a concern at this stage, Preminger said.
That said, Preminger offers a disclaimer: some errors are to be expected.
“This is something where it gives you a draft, as your paralegal or associate would do,” Preminger said. “And then you get to look it over, you edit it, you check it, you make your additions, and then you would file it.”
Preminger’s fielded a lot of questions from his peers who worry about the pitfalls of relying on AI, he said, and was recently invited to speak about LitLaw at a legal conference this fall. He’s tried to reassure lawyers by vowing to launch LitLaw with care and continually monitor for problems.
“I think, to some extent, it takes understanding this is a new technology and you can have mishaps. Everybody knows now about hallucinations, which they didn’t before,” Preminger said. “But, I think it’s going to be kind of up to us to make sure to roll this out in a really careful way and make sure we don’t have those kinds of traps.”
“This is a dedicated server for litigation and it’s kind of a closed playground to some extent,” Preminger added. “But also, it’s going to be up to the lawyer to check everything. A lawyer does need to read this all over, make sure the definitions are correct, make sure they have the right things going on.”