Clayton Utz adopts Lexis+ AI for legal research; Holding Redlich reports one research task cut from 4.5 hours to 30 minutes

Lexis+ AI, from LexisNexis, gives lawyers access to a database of legislation, case law and commentary through a conversational AI assistant. The article cites a Stanford University study finding that generative AI tools designed for the legal industry make up false or misleading information between 17% and 33% of the time. LexisNexis's Asia Pacific managing director names client confidentiality and references to non-existent statutes or case law as the two greatest risks of applying generative AI to the legal sector.

Clayton Utz and Holding Redlich became the first Australian law firms to adopt Lexis+ AI, a generative AI solution for legal research. The article says it is built on Anthropic's Claude 2 and OpenAI's GPT-4 and trained on LexisNexis' repository of 1.23 million court opinions, statutes, filings, and secondary materials. It offers conversational search, summarisation and document drafting, and users are required to review and adjust AI-generated drafts. LexisNexis says prompts are encrypted, the model does not learn from them, and they are deleted. The article says that after generating a court document, Lexis+ AI validates its citations against its entire database of primary and analytical content. Clayton Utz helped test and develop the product and was adopting it for day-to-day use.

Keren Smith, Chief Knowledge Officer at Holding Redlich, said that one of its lawyers spent four and a half hours on a legal research task using traditional methods and replicated the same task in 30 minutes using Lexis+ AI. This was a single task at Holding Redlich. Smith said the case summary feature "significantly expedites our research, allowing lawyers to quickly identify relevant cases and avoid irrelevant ones," and that lawyers "still need to read cases and apply the same standard of due diligence such as checking accuracy of footnotes and citations as they would with traditional legal research." Clayton Utz Chief Executive Partner Emma Covacevich said products like Lexis+AI, "when deployed with appropriate use guardrails, offer efficiencies that haven't previously been possible. And while no AI solution is perfect, the rapid advancements in this space are exciting."

• A Stanford University study cited in the article found false or misleading information from legal-industry AI tools 17% to 33% of the time, and the article says Lexis+ AI validates the citations in a generated court document against its database.
• LexisNexis says prompts are encrypted, not learned from by the model, and deleted.
• Clayton Utz helped test and develop the product before adopting it for day-to-day use.
• Holding Redlich's Smith says lawyers still need to read cases and check footnotes and citations.