AI Misconduct.com

The daily rap sheet of the machine age.

Updated Nov 2, 2026
4,173 items in the archive
RSS feed

Tuesday, May 6, 2025

11 items

Lawyers' five fake case citations earn wasted costs, regulator referral

Who did it: HUMAN Legal Futures

Mr Justice Ritchie ordered barrister Sarah Forey and Haringey Law Centre to pay £2,000 each in wasted costs, and their regulators to be sent his judgment, over five fake cases cited in a housing claim against Haringey. The council's counsel suggested AI. The judge could make no finding on that, but rejected Ms Forey's drag-and-drop explanation.

curl flags AI-generated fake bug bounty report on HackerOne

Who did it: HUMAN Socket

The curl project flagged HackerOne report H1 #3125832 as AI-generated slop citing nonexistent functions and fake patches, according to security researcher Harry Sintonen. The article says the reporter, linked to the @evilginx account, appears to have tried similar tactics elsewhere and received payouts in some cases; Open Collective and Python Software Foundation's Seth Larson report similar AI-generated submissions.

Grok answers 'remove her clothes' requests on women's photos on X

Who did it: HYBRID Business and Human Rights Centre

Anyone on X can reply to a photo a woman has posted asking Elon Musk's Grok to "remove her clothes", and the chatbot answers in-thread with an image of her in a bikini or lingerie, 404 Media reports after a Bellingcat researcher flagged it. Grok balks at full nudity on consent grounds, once after already putting a woman in her underwear.

Two law firms ordered to pay $31,100 over AI-invented citations

Who did it: HYBRID AI Hallucination Cases database

Special master Michael Wilner struck the plaintiff's briefs in Lacey v. State Farm after about nine of 27 citations proved wrong and at least two cases nonexistent, traced to Ellis George lawyer Trent Copeland's use of CoCounsel, Westlaw's AI tool and Google Gemini. Ellis George and K&L Gates must pay the defense $31,100; the "fixed" brief kept a half-dozen AI errors.

Ontario judge orders lawyer to show cause over suspect citations

Who did it: HYBRID canlii.org

Ontario Superior Court Justice FL Myers ordered lawyer Jisuh Lee to show cause why she should not be cited for contempt after her factum linked to unrelated or unfindable cases and cited one ruling for the opposite of what it held. The judge said the factum may have been AI-generated; asked, she said she would check with her clerk.

MrDeepFakes shuts down after losing a critical service provider

Who did it: HUMAN NPR

MrDeepFakes, a site for nonconsensual sexual deepfakes mostly of celebrities, has closed, saying a "critical service provider" cut it off permanently and data loss made continuing impossible. It says it will not relaunch. A 2024 Stanford and UC San Diego study counted 43K videos of 3.8K people as of November 2023, watched more than 1.5B times.

Fake ComfyUI extension author pleads guilty to hacking Disney employee

Who did it: HUMAN Ars Technica

Ryan Mitchell Kramer, 25, pleaded guilty to accessing a computer and obtaining information and to threatening to damage a protected computer, after a Disney employee installed his malicious ComfyUI_LLMVISION extension in April 2024. He then downloaded roughly 1.1 terabytes from thousands of private Disney Slack channels and later leaked it, along with the employee's bank and medical details.

Russia-aligned Operation Overload clones experts' voices for fake news videos

Who did it: HYBRID Institute for Strategic Dialogue

ISD says Russia-aligned Operation Overload impersonated more than 80 organisations across 135 pieces of content between January and March 2025, often adding AI-generated voices to real footage of academics, journalists and police. One video falsely claiming USAID paid celebrities to visit Ukraine drew over 4 million views; most other engagement came from bots.

Knesset panel weighs doctors' ChatGPT use; advocate cites triple dose

Who did it: HYBRID כיפה

The Knesset health committee discussed doctors' growing use of AI after the chair of the group Lema'anchem appealed to the Health Ministry, warning of unregulated reliance on ChatGPT. Among the cases the chair cited: a patient prescribed a drug at three times the correct dose after the doctor relied on a figure from ChatGPT.

Tesla on Autopilot rams parked van in driveway; driver ticketed

Who did it: HYBRID ABC7 New York

A Tesla that police say was on Autopilot accelerated into a South Brunswick, New Jersey, driveway and pushed a parked minivan into a garage; the van's owner, planting seedlings nearby, escaped injury. Police believe the car mistook the driveway for a street while avoiding an oncoming SUV; the Tesla's owner was ticketed for careless driving.

Judge warns plaintiff suing Take-Two over unfindable cases and quotes

Who did it: HYBRID AI Hallucination Cases database

Take-Two Interactive told a Colorado federal court that pro se plaintiff Colton John Harris presented quotations absent from the cited cases, or cases that couldn't be found, noting courts' warnings about generative AI. Denying Harris's reconsideration motion, Chief Judge Philip A. Brimmer warned such fictions may draw Rule 11 sanctions; Harris had withdrawn the citations or said he wasn't quoting directly.