Prosecutors have charged Matthew Derrick Hudson with wire fraud over a scheme that raised over $100 million. Hudson is the founder and former CEO of Invenia Technical Computing Corporation, which claimed to apply AI to North American energy markets. The complaint alleges he sent investors an audit report claiming about $218 million CAD in cash; the real audit showed about $6 million CAD.
Boies Schiller Flexner partner John Kucera took responsibility in a Sept. 19 declaration for "material citation errors" in an AI-assisted brief for plaintiffs suing the Church of Scientology and Danny Masterson. Church lawyers had flagged the errors, saying one cited case "is completely made up." Kucera said he failed to proofread and cite-check it, and he is seeking to file a corrected brief.
California's 2nd District Court of Appeal fined Los Angeles-area attorney Amir Mostafavi $10,000 on Sept. 12, finding that 21 of 23 quotes from cases cited in his opening brief, generated with ChatGPT, were made up. Mostafavi told the court he did not read the AI's text before filing, and told CalMatters he didn't know it would add citations.
Belgian digital minister Vanessa Matz told a parliamentary justice committee that Tesla's Sentry Mode, which films a parked car's surroundings with its exterior cameras, breaches Belgian and European law. She cited a 2007 law barring private individuals from filming public space and GDPR fines of up to €20 million for not informing those filmed; nearly 21,000 Teslas in Belgium are potentially affected.
Former and current students sued Lawrence Public Schools in Kansas in August to stop its use of Gaggle, an AI tool scanning student documents and emails, alleging the monitoring is unconstitutional and prone to misfires. Students say it deleted an art student's photo of girls in tank tops after mistakenly flagging it as child pornography, and intercepted the school's reply to a records request.
Scammers posing as "General Hospital" actor Steve Burton courted a woman via Facebook and WhatsApp, sending a video experts say is a deepfake of him professing love, the Los Angeles Times reports. Her family says she sent them $81,000 and, at the scammer's urging, sold her home; the family's lawsuit to stop the transfer is pending.
YouTube channel "Chroniques Quotidiennes" posted an AI-generated video on September 6 claiming Emmanuel Macron's nephew was arrested in Burkina Faso with 3.7 million euros in cash; many users shared it as authentic. FRANCE 24 Observers found the channel, whose profile calls its content "entirely fictional", has published 344 pro-Traoré videos since February 26, 54 targeting Macron.
Scammers used AI to make Facebook videos in which former Polish first lady Jolanta Kwaśniewska appears to promote a supposed government programme paying 3,000 zł a month, aimed at people aged 50 and over. An older woman who lost a large sum phoned Kwaśniewska's daughter; Kwaśniewska told TVN24 she never encouraged any investments, and the matter was reported to police.
The Baldwin County, Alabama, Sheriff's Office arrested Brandon Michael Fry on Sept. 23 on one count of sexual abuse of a child under 12 and four counts of possession of child pornography. Investigators say his computer held several hundred AI-generated abuse images made from pictures of children he knew, and that he admitted knowingly possessing the material.
A Houston man says he lost $15,000 to scammers who used AI to mimic his son, supposedly arrested for DUI, then had a fake Harris County public defender collect "bond" through four Coinme crypto deposits. Coinme says the funds were removed within minutes, and the constable's office says such calls usually can't be traced.
Who did it: HYBRID AI Hallucination Cases database
Ruling Sept. 24 in a sex-discrimination suit against the Mississippi Department of Rehabilitation Services, U.S. District Judge Daniel P. Jordan III said the plaintiff's initial reconsideration briefs violated Rule 11; counsel had admitted they contained false AI-generated content. He said self-reporting was right but noted defendants flagged incorrect authority in February and counsel waited seven months; reconsideration was denied.
The Queensland Audit Office found the Department of Transport and Main Roads never carried out required ethical risk checks on its AI-powered phone and seatbelt cameras, which in 2024 made more than 208 million assessments, leading to roughly 114,000 fines bringing in over $137 million. Transport Minister Brent Mickelberg said the department has begun implementing the auditors' recommendations.
The Dutch Data Protection Authority (AP) says it is very concerned about LinkedIn's plan to train its AI from November 3, 2025 on European users' profile data and public posts dating back to 2003, with the setting on by default. It urged users to opt out; it has received complaints, which it is handling with Ireland's DPC, LinkedIn's lead regulator.
Microsoft says a credential phishing campaign hid its payload in SVG files dressed up as PDFs, encoding malicious code as business terms like "revenue" and "shares" behind an invisible fake analytics dashboard. Its Security Copilot judged the code likely LLM-generated, citing complexity and verbosity unusual in human-written scripts; Microsoft says it detected and blocked the campaign.
