In a recent concurrence from the 11th Circuit’s decision in Snell v. United Specialty Ins. Co., Judge Kevin Newsom explored the potential for AI-powered large language models (LLMs) to aid in legal interpretation. While the case itself concerned an insurance dispute over whether a ground-level trampoline project counted as “landscaping” under the policy terms, Judge Newsom’s concurrence delved into broader implications of using LLMs in law.
Judge Newsom, after personally testing LLMs to define “landscaping,” recorded his thoughts on the responses and the utility of LLMs in understanding legal texts based on their massive training data derived from diverse internet sources. He outlined five primary benefits of using LLMs, including their ability to process vast amounts of data, understand context, accessibility, transparency in methodology, and their advantages over traditional empirical interpretive methods like dictionaries and corpus linguistics.
However, Newsom also considered potential drawbacks, such as the propensity of LLMs to produce “hallucinations” or erroneous outputs, their limited capture of offline speech which might exclude underrepresented populations, the possibility of manipulation by legal professionals, and the dystopian prospect of overly relying on AI in judicial decisions. Despite these concerns, he concluded that the benefits of integrating LLMs into legal interpretation outweigh these risks, provided the models are used judiciously.
This concurrence, as highlighted in the GT Alert from Greenberg Traurig, LLP, not only adds to the ongoing discourse on the intersection of AI and law but also suggests practical steps for incorporating AI tools in legal processes, emphasizing transparency, varied input prompts, and a nuanced understanding of language usage over time. Judge Newsom’s insights contribute to the evolving conversation about how AI can enhance legal interpretations without replacing the essential human judgment required in the judiciary.