Patent lawyers slapped with $12,000 fine for AI-generated submissions: level of hallucinations in court filings “staggering”, says judge

Context:

• Artificial intelligence use in patent drafting and court filing is on the rise – and while it has become an incredibly useful tool for many attorneys and courts, its use also comes with pitfalls. ai fray reported on some of the risks that come with over-reliance on AI last June, including through hallucinations (fabricated concepts, fake prior art, or unworkable ideas masked by technical jargon), according to Dr.-Ing. Robert Klinski, a patent attorney and founder of Germany-based firm PATENTSHIP (June 20, 2025 ip fray article).
• Last year, licensing firm Lexos Media IP LLC sued online retailer Overstock.com over the alleged infringement of three patents related to modifying cursor image displays on websites in the United States District Court for the District of Kansas. Counsel for the plaintiff submitted briefs containing case citations, quotations, and representations of legal authority fabricated by artificial intelligence – hallucinations produced by Generative AI tools like ChatGPT. In December, United States District Judge Julie Robinson of the District of Kansas ordered the attorneys – from Fisher Patterson Sayler & Smith, Buether Joe & Counselors, and Seth Law – to show why they should not be fined for the misrepresentations under Federal Rule of Civil Procedure 11.

What’s new: Judge Robinson has fined Seth Law’s Sandeep Seth $5,000, who admitted to using ChatGPT without verification for submitting documents, while she fined Kenneth Kula and Christopher Joe of Buether Joe & Counselors $3,000 each, and local counsel David Cooper of Fisher, Patterson, Sayler & Smith a further $1,000. While Mr. Seth was the only one to use GenAI to file case materials, the judge held all four lawyers responsible for failing to check them. Judge Robinson also ordered Mr. Seth to provide state disciplinary authorities a copy of the ruling and certify what his firm will do to prevent a similar situation in the future. She wrote:

“A reasonably competent attorney filing documents in court should be aware of the pronounced, well-publicized risks of using unverified generative AI for legal research and the ethical obligations associated with signing a court filing without checking it for accuracy.”

Direct impact: In a statement, Mr. Seth said that the case has been “an embarrassing lesson”. “Firms should not use AI as a tool in any capacity without strict policies in place in order to avoid errors,” he added.

Wider ramifications: While this appears to be one of the first patent-related AI hallucination cases, in 2025 there were several cases around the world (non-IP) in which attorneys were fined for hallucinated legal material. In her decision, Judge Robinson also wrote that the amount of case law that has “erupted” over the last few years due to attorneys’ reliance on unverified GenAI research, often generating hallucinated legal authority, is “staggering”. In December, a judge in California fined Robert Carey and Celeste Boyd of Hagens Berman $13,000. In September, a judge in Puerto Rico sanctioned two lawyers $24,400 for submitting court filings in a FIFA suit that included at least 55 erroneous or non-existent case citations. And, in early 2025, attorneys from the firm Morgan & Morgan (including Rudwin Ayala) in a Walmart-related case were hit with a collective $5,000 fine after using AI that hallucinated case law in filings.