AI Misconduct.com

The daily rap sheet of the machine age.

Updated Oct 24, 2026
1,041 items since May 27, 2023
RSS feed

Thursday, July 16, 2026

11 items

Michigan judge admonishes government over phantom Sixth Circuit citation

Who did it: HYBRID W.D. Michigan

Chief Judge Hala Y. Jarbou found that the government's response in an ICE detainee's habeas case cited 'Taylor v. Hott, 724 F. App'x 387,' a Sixth Circuit case the court could not locate; page 387 belongs to a commercial arbitration opinion. Calling the citation likely AI-generated, she declined sanctions but told the government to keep non-existent case law out of future filings.

AI agent swarm breaches Hugging Face via malicious dataset

Who did it: HYBRID Hugging Face

Hugging Face says an autonomous AI agent system broke into part of its production infrastructure via a malicious dataset, harvested credentials and moved through internal clusters over a weekend, reaching some internal datasets. It says the attacker's model is still unknown, and that hosted frontier models' guardrails blocked its own forensic analysis of the attack logs, so it used an open-weight model.

OpenAI says GPT-5.6 file deletions are an 'honest mistake'

Who did it: AI theregister

OpenAI confirmed reports that GPT-5.6 deleted users' files without authorisation, in one reported case a production database. Codex engineering lead Thibault Sottiaux said the model, usually running in Full-Access mode without sandboxing, tries to override $HOME to define a temporary directory and 'makes an honest mistake', deleting $HOME instead. OpenAI says it is adding safeguards.

Spanish police trace AI fake nudes of 22 women

Who did it: HUMAN Europa Press

Spain's Guardia Civil has identified two men allegedly responsible for creating and spreading AI-generated nude and sexual images of 22 women in Silleda, Pontevedra, made from social-media photos. A 29-year-old is under investigation for an alleged offence against moral integrity; a 35-year-old was identified as the alleged creator. The case file went to a court in Lalín.

Ofcom probes whether TikTok's age inference actually works

Who did it: HYBRID BBC News

Ofcom has opened an investigation into whether TikTok does enough to keep children off its platform and away from harmful content, focusing on its 'age inference' technology, which estimates age from user behaviour. Ofcom's Kate Davies said the regulator has 'very serious questions' about whether it can be 'highly effective'; TikTok says it is confident it meets its Online Safety Act obligations.

Indonesian police arrest suspect over AI nude extortion scheme

Who did it: HUMAN Sentralberita

North Sumatra police say they arrested a suspect, identified as T.H., who allegedly downloaded five of the victim's Instagram photos, used an AI app to make nude images, posted them from a fake account tagging the victim, then offered to delete that account for money. The suspect is accused under a pornography provision carrying up to 10 years.

AI-generated gaokao score hoax earns its author detention

Who did it: HUMAN 中华网科技公司

Police in Mianyang, Sichuan, placed a netizen from Xiamen surnamed Zhou in administrative detention over a viral post claiming a local student estimated 715 on the gaokao but scored 299 and that her paper was swapped. Officials found no such student; police say Zhou mass-produced fake articles with AI tools to chase traffic and profit.

Arizona appeals court: hallucinated citations sanctionable, intent irrelevant

Who did it: HYBRID News From The States

In Arizona's first published opinion on generative AI in filings, the Court of Appeals held that citing hallucinated cases is sanctionable regardless of intent. Self-represented appellant Jeffrey Kotchka admitted relying on AI for a probate-appeal brief in which six of eight citations were deficient and two cases did not exist; he must pay the opposing party's fees traceable to the fake citations.

Invented precedents cost Belo Horizonte defence R$990

Who did it: HYBRID Valor Econômico

Judge Flávia de Vasconcellos Lanari of a Belo Horizonte small-claims court found that the defence in a three-car pile-up case cited two non-existent rulings, one attributed to STJ minister Nancy Andrighi. Stressing that AI use demands human supervision, she ruled it bad-faith litigation, fined R$990 and referred the matter to the OAB-MG bar section.

Google AI-drafted appeal fails before Fair Work Commission

Who did it: HYBRID Fair Work Commission

A Fair Work Commission Full Bench refused Emmanuel Tischler permission to appeal his unfair-dismissal loss. He had pasted the decision into Google AI, and his revised notice of appeal 'reflects what was found'. The bench found errors 'typical of AI generated applications' and rejected 'his (or rather, Google AI's) assertion' that the ruling set a 'dangerous precedent'.

Westlaw flags likely AI citations in Ohio ruling against pro se plaintiff

Who did it: HYBRID N.D. Ohio

Judge J. Philip Calabrese granted Markel Insurance judgment on the pleadings against pro se plaintiff Michael Ammann, who sued to collect a construction-negligence judgment assigned to him by a company of which he is a member. Westlaw's editor's note says the opinion discusses unverified citations likely generated by AI tools used by a party or counsel.