AI Misconduct.com

The daily rap sheet of the machine age.

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

Thursday, February 19, 2026

8 items

Lawsuit says ChatGPT convinced student he was an oracle

Who did it: AI Ars Technica

A Georgia college student has sued OpenAI in San Diego Superior Court. The suit alleges GPT-4o convinced him he was an oracle, compared him to Jesus and Harriet Tubman, and set him a process of unplugging from everyone but ChatGPT. It says he was hospitalized for a week and diagnosed with bipolar disorder; it is the 11th known suit of its kind.

Grok hands out porn performer Siri Dahl's legal name unprompted

Who did it: AI 404 Media

Asked on X who appeared in a mislabelled porn clip, xAI's Grok volunteered performer Siri Dahl's legal name and birth date, which she says she had spent thousands of dollars on data-removal services to keep private. Almost instantly, she says, harassers opened Facebook accounts and posted stolen clips under her real name.

Police say suspect in Seoul motel deaths quizzed ChatGPT

Who did it: HUMAN The Korea Herald

Seoul police say a woman in her 20s surnamed Kim asked ChatGPT "Could it be fatal?" about mixing sleeping pills and alcohol before allegedly killing two men with drinks laced with her prescribed benzodiazepines in Gangbuk-gu motels. Citing the queries as evidence of intent, police referred her to prosecutors on murder charges; she admits spiking the drinks but denies meaning to kill.

Fidesz campaign video stages AI-made execution of a father

Who did it: HUMAN Reuters

Hungary's ruling Fidesz posted a 33-second campaign video, made with the help of Google's AI models according to Reuters, showing a girl weeping while her father, a blindfolded soldier in Hungarian uniform, is shot, captioned "Fidesz is the safe choice!" Opposition leader Peter Magyar called it "sickening, unforgivable and deeply outrageous"; Orban's chief of staff did not deny AI use.

Gordon Rees accused of fabricated citations, months after AI rebuke

Who did it: HUMAN Above the Law

In Huynh v. Redis Labs, plaintiff's counsel Bach Mili claims a Gordon Rees opposition brief reverses holdings and fabricates quotations, calling it "sanctionable on its face". The reply doesn't blame AI outright, but notes the same partner drew a December 3, 2025 reprimand for AI-hallucinated citations, months after the firm's August benchslap in a bankruptcy case.

ESET finds Android malware that asks Gemini where to tap

Who did it: HUMAN The Register

ESET says PromptSpy, Android malware apparently built by Chinese speakers, sends Google's Gemini an XML dump of the screen and follows its JSON instructions to keep itself pinned in recent apps, while intercepting PINs and blocking uninstalls with invisible overlays. It has not appeared in ESET telemetry, but a distribution domain was likely imitating Chase Bank.

Former Texas student arrested over AI nude images of students, staff

Who did it: HUMAN wfaa.com

A former high school student in North Texas has been arrested after allegedly photographing students and staff in their district and neighboring ones and running the pictures through an AI platform to create sexually explicit images, the district's superintendent says. The Texas Rangers and local police, who have the suspect's phone, are investigating.

Galgotias University exits AI summit over Chinese-made robot dog

Who did it: HUMAN The Times of India

Galgotias University vacated its stall at the India AI Impact Summit after a professor said on camera its robot dog "Orion" "has been developed by the Centre of Excellence" and online users identified it as a Chinese-made Unitree Go2. After a day of shifting explanations, it apologized, calling its representative "ill-informed" and insisting there was "no institutional intent to misrepresent".