insta_photos // Shutterstock
1 / 2The legal hallucination problem
Understanding AI's limitations could be the difference between winning and losing a case for people considering using it for legal help. A New York attorney learned this when he used ChatGPT for legal research in an injury case. The federal judge overseeing the lawsuit found that the AI had made up fake case citations and quotes, claiming these made-up sources could be found in major legal databases. If trained lawyers can be misled by AI's confident-sounding but false information, the risks for people handling their own legal matters could be greater.
These "hallucinations"—the term used to describe convincing but false information that AI makes up—pose real risks in legal situations. According to legal experts, people should be cautious about using AI to interpret contracts, analyze their legal rights, or prepare court documents as the AI might invent legal principles or misinterpret important details that could harm their case. This happened in the United Kingdom, where a person representing themselves inadvertently made entirely fake submissions to the court based on ChatGPT-generated advice.
The problems with AI's accuracy are concerning for people seeking legal help. Most legal documents are kept behind expensive paywalls, meaning AI systems often lack access to important legal information. For example, a 2022 study of one of the largest databases offering free access to British and Irish legal judgments found that only half of court decisions can be found on the site. This means when someone uses an AI tool to help understand their legal situation or prepare documents, they have no way of knowing if it is working with complete or accurate information.
Legal organizations have raised several specific warnings about using AI for legal help. The State Bar of California warned that AI can sound confident even when giving wrong advice, making it hard for users to spot mistakes. LexisNexis, a leading provider of legal research and analytics tools, emphasized that AI should never make final legal decisions without human review. Meanwhile, Clio, a major legal practice management software company, warned that AI systems might give biased advice if they learned from biased historical data.







