Courts and regulators are stepping up penalties as generative AI errors show up in legal filings. High-profile incidents last year highlighted the risk: lawyers for MyPillow CEO Mike Lindell were each fined for submitting briefs that relied on AI and included fabricated case citations. That episode proved far from isolated.
A researcher who tracks court punishments for AI-related errors has recorded more than 1,200 instances worldwide, roughly 800 in U.S. courts, and the tally keeps growing. Observers say the problem stems from AI tools that appear authoritative but sometimes invent or misstate facts and authorities.
Sanctions are getting larger. A federal judge recently ordered an Oregon attorney to pay roughly 109,700 dollars in sanctions and costs after filing briefs with AI-driven inaccuracies, possibly setting a new high-water mark. State supreme courts have also confronted similar situations: Nebraska questioned an Omaha lawyer about a brief that cited nonexistent cases and referred him for discipline after he denied using AI, and Georgia faced a comparable matter.
Law librarians and legal educators are pushing for training to address the gap. An associate dean at the University of Washington law library is developing optional ethics instruction on generative AI for students, emphasizing an enduring rule: lawyers remain responsible for the accuracy of everything they file, however it was produced. Practitioners must read, verify, and understand any cases or authorities an AI suggests before citing them.
Some courts are going further by adopting disclosure rules that require attorneys to state when and how AI tools were used in a filing. Proponents say labeling signals which documents warrant extra scrutiny and helps distinguish human judgment from automated assistance. Critics counter that such rules will soon be unworkable because AI is being embedded across legal software; when everything is AI assisted, disclosure could become meaningless and routinely ignored.
Commentators also warn about more ambitious, so-called agentic systems that promise end-to-end completion of legal tasks. Those systems can obscure the intermediate steps where errors originate, increasing the chance that even careful lawyers will miss mistakes. At the same time, AI-driven speed threatens the traditional billable-hours model. If firms accept lower fees or adopt new billing practices as tasks become faster, time pressure could encourage reliance on AI drafts without adequate review. Observers worry that younger lawyers raised on always-on AI may be less inclined to pause and think critically.
Still, some experts reject the notion that AI will replace lawyers wholesale. The prediction is instead that lawyers who learn to use generative AI responsibly will outperform those who do not.
AI is also being targeted in litigation. In March, an insurance company sued a prominent AI provider in federal court, alleging that flawed AI legal advice led to frivolous actions and accusing the company of practicing law without a license; the provider called the complaint baseless.
For now, the legal profession is adapting. Courts are enforcing existing duties of competence and candor, law schools and firms are developing training and policies, and debate continues over disclosure, supervision, and the proper role of AI in legal work. The consistent takeaway: generative AI can increase efficiency but attorneys must verify and take responsibility for the content they submit to courts.