Source: State Bar of Wisconsin

Generative artificial intelligence (GAI), particularly large language models (LLMs), is increasingly used in legal practice due to its ability to outline legal arguments, summarize complex documents, and assist with various tasks. However, these systems produce “hallucinations,” which are outputs that appear confident and professional but are actually incorrect, unsupported, or misleading. Hallucinations arise not from system glitches but from the predictive nature of LLMs, which generate text based on likely next words without inherent legal accuracy.

There are several types of AI hallucinations relevant to lawyers. One is fabricated authority, where the AI invents case names, citations, or quotations that do not exist. In Mata v. Avianca Inc., a federal court sanctioned lawyers who submitted filings with fabricated judicial opinions and fake quotations generated by ChatGPT, highlighting the ongoing gatekeeping responsibility of attorneys to verify accuracy under court rules. Failure to identify AI hallucinations may expose lawyers to sanctions, disciplinary action, and risks for their clients and reputations.