AI Misconduct.com

The daily rap sheet of the machine age.

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

Friday, April 4, 2025

10 items

Grok 3 'wrote' climate-denial paper, its co-authors say

Who did it: HYBRID ETBrandEquity.com

A paper titled "A Critical Reassessment of the Anthropogenic CO2-Global Warming Hypothesis", which says Grok 3 "wrote the entire manuscript", is being promoted by climate change deniers; experts told AFP it relies on contested references, including papers by co-author Willie Soon and by Hermann Harde, the journal's editor-in-chief. It went from submission to publication within 12 days.

StrongRoom AI collapses weeks after raise; investor alleges fraud

Who did it: HUMAN Capital Brief

Weeks after Sydney venture firm EVP led a $10 million investment into pharmacy software startup StrongRoom AI, its CFO resigned and the company collapsed into administration. EVP called in police and a Federal Court froze the startup's assets; unsealed court documents set out what EVP alleges was a "deliberate fraud". The founders have not spoken publicly.

Study suggests GPT-4 memorized copyrighted books and news

Who did it: HYBRID TechCrunch

Researchers at the University of Washington, the University of Copenhagen and Stanford, testing whether OpenAI models could guess masked "high-surprisal" words, found GPT-4 showed signs of having memorised portions of popular fiction books, including copyrighted e-books in the BookMIA dataset, and, at a lower rate, New York Times articles, lending credence to infringement allegations against OpenAI.

Essex man jailed for five years over AI deepfake porn

Who did it: HUMAN BBC

Brandon Tyler, 26, of Braintree, was jailed for five years at Chelmsford Crown Court after admitting 18 counts of causing harassment without violence and 15 of sharing intimate images. The court heard he used AI to manipulate social media photos of 20 women he knew, targeting them in 173 posts between March 2023 and May 2024 on a forum glorifying "rape culture".

Litigant's undisclosed AI avatar argues New York appeal, briefly

Who did it: HUMAN Associated Press

Jerome Dewald, representing himself in an employment dispute, played a video on March 26 in which an AI-generated avatar began arguing his appeal before New York's Appellate Division, First Judicial Department. Justice Sallie Manzanet-Daniels had it shut off, saying "I don't appreciate being misled." Dewald later apologised to the court, saying he hadn't intended any harm.

ACLU says Cleveland police hid facial recognition from warrant judge

Who did it: HUMAN ACLU of Ohio

The ACLU, the ACLU of Ohio and the NACDL have filed an amicus brief with Ohio's 8th District Court of Appeals backing a trial judge who suppressed evidence in a murder case. They say Cleveland police concealed from the warrant judge their reliance on a Clearview AI facial recognition search, which police testified returned "multiple photos of multiple people".

Delaware judge warns pro se litigant over phantom case quotes

Who did it: HYBRID AI Hallucination Cases database

Delaware Vice Chancellor Lori W. Will denied with prejudice a discovery motion by pro se petitioner Daniel Jaiyong An against Archblock, finding cited cases that do not support the propositions and quotes the court could find in no case anywhere. The judge suspected generative AI may have "hallucinated"; An was warned, not sanctioned, and future filings must include a certification on GenAI use.

Viral Romanian 'poor boy' Facebook story was AI-made, Edupedu says

Who did it: HYBRID Edupedu.ro

A Facebook page called "Viață de Românaș" posted a tear-jerking tale of "Mihăiță", a poor Bucharest boy at a traffic light, that drew over 22,000 likes in under a day. Edupedu.ro says the boy, his family and the story are invented, the image AI-generated and the text very probably written with a model like ChatGPT.

TikTok deepfake has Uhuru Kenyatta announcing 2027 presidential bid

Who did it: HUMAN africacheck.org

A 12-second TikTok video viewed more than 820,000 times appears to show former Kenyan president Uhuru Kenyatta announcing a 2027 presidential run. Fact-checkers say it is a deepfake, apparently built from old footage from his presidency with AI-generated audio; Kenyatta has already served the two terms the constitution allows.