AI keeps inventing fake cases. Lawyers keep citing them

The Alabama Supreme Court sanctioned an attorney in April after they filed legal briefs containing AI-generated citations to nonexistent cases. Despite a promise to correct the error, the lawyer continued submitting fabricated precedents in subsequent filings, prompting further judicial action. Another attorney was also sanctioned the same week for persisting with AI-hallucinated material after warnings. A database maintained by Damien Charlotin, a senior research fellow at HEC Paris, documents over 1,400 court decisions addressing AI errors in legal filings over the past three years. Charlotin notes that while the rate of reported cases has slowed from exponential growth, courts still issue around 350–400 rulings per quarter on AI-related inaccuracies. He has also developed an AI-powered reference checker called Pelaikan to help mitigate such errors. Beyond legal proceedings, AI-generated inaccuracies have affected journalists, software developers, academic researchers, and government consultants. In May, *The New York Times* reported that the author of *The Future of Truth*, a book on AI’s impact on discourse, admitted to including multiple fabricated or misattributed quotes produced by AI tools. The trend reveals a persistent issue: professionals continue trusting AI despite knowing its fallibility. This misplaced confidence has led to dismissed legal appeals, attorney fines, and even job losses. A February study found participants were more likely to trust AI-generated guidance—even when it was only 50% accurate—compared to human-provided advice. Experts warn that as AI becomes more embedded in professional work, the stakes will rise. Alan Wagner, an associate professor at Pennsylvania State University, explains that humans often overestimate machines’ knowledge and reliability, assuming they are infallible. The realistic yet false outputs of AI further complicate this trust issue, making its errors uniquely believable to users.