Involving LLMs in legal processes is risky

An invited paper

Abstract

The availability of large language models (LLM), a form of AI correspondence agent, on the Internet raises questions about their possible use in legal processes. For they seem often to introduce into their output arbitrary assertions which are untrue. This is ‘new’ behaviour (a Google search, for example, returns correct information on a semantically similar query). I give examples in which use of an LLM is both tempting and could subvert a legal process. 

Index words: large language model; LLM; artificial intelligence; AI; ChatGPT; law; legal processes; lying; r-lying