A federal magistrate judge in Wisconsin fined attorney Jennifer DeMaster $4,000 and co-counsel Chris Kachouroff $500 over misrepresented case law, a non-existent case and factual misstatements in summary judgment filings against the City of Sheboygan. The judge noted the pair, sanctioned for similar conduct in Coomer v. Lindell, had first-hand knowledge from that case that AI could fabricate caselaw.
Recorded Future's Insikt Group says it has found at least 200 new fake news sites since January attributed to the Russian network CopyCop, many posing as fictional local outlets in the US, France, Canada and Norway, with self-hosted uncensored LLMs based on Meta's Llama 3 generating at least some stories. Researchers claim John Mark Dougan likely operates the sites.
KBS, MBC and SBS faced Naver at the Seoul Central District Court's first hearing of their suit alleging it used their news without permission to train HyperCLOVA and HyperCLOVA X, seeking 200 million won each. They cited Naver's AI confirming it had; Naver argued its content terms permitted use and the model tends to answer as questioners intend.
Entertainer Chen Meifeng said on social media that someone had used AI to fake her appearance and voice in online ads selling skincare products, posting screenshots of the ads. Saying she was 'so angry', she urged fans to warn older relatives about scams and said AI-generated video and voice should be covered by law.
Tesla's crash report to the federal government lists three crashes involving its Austin robotaxis, all dated July 1, ten days after the pilot launched; one, hitting a stationary object at 8mph, caused minor injuries. Tesla redacted much of the data, claiming it reveals proprietary information, and the reports say no safety driver was present, which the writer calls very unlikely.
A New Jersey federal magistrate judge fined attorney Sukjin Henry Cho $3,000 after his reply brief for OTG New York cited nonexistent cases and fabricated propositions from generative AI. Cho admitted using AI without verifying the output; the court struck the reply and ordered him to self-report to his state bars and serve the order on his client.
An appeals court in General Roca, Río Negro, found nonexistent rulings cited in briefs: 20 from the plaintiff's lawyer, Sergio D'agnillo, and six from Nicolás Gatti, lawyer for co-defendant DMX2, in a car-defect suit against Peugeot Citroën Argentina. Saying AI use was very probable, it denied both lawyers fees for the appeal, issued a severe warning and notified the local bar association.
Google quietly added a 'homework help' button to Chrome on Sept. 2 that appeared on course websites and, educators demonstrated, served Google Lens 'AI Overview' answers during quizzes. Several days after The Washington Post contacted Google, the company said it had temporarily paused the button, but did not commit to keeping it off.
Justice Devin P. Cohen of the New York Supreme Court in Kings County denied a motion by Achievement First-linked defendants in a construction-site damages suit and found their reply papers contained what appears to be a fictitious citation with a fabricated holding. Unconvinced it was a mere scrivener's error, he set an October 1, 2025 hearing on costs and/or sanctions.
The Ghent Court of Appeal found a brief for a mother in a family case cited untraceable case law, misquoted statutes including the Judicial Code, and quoted entirely fictitious Belgian legal scholarship. Asked about AI, her lawyer said a staff member drafted it and the sources went unchecked; the court called such conduct wholly unacceptable and disregarded the sources.
Dismissing a self-represented plaintiff's wrongful-repossession suit against Stellantis Financial Services, U.S. District Judge Roderick C. Young in Virginia noted that many if not all of the plaintiff's citations were inaccurate, suggesting AI drafting. He proposed that the time may be near for an exception to the rule giving pro se pleadings liberal construction when litigants rely on AI.
