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, June 18, 2026

13 items

aespa deepfake seller gets two years and six months

Who did it: HUMAN The Kyunghyang Shinmun

The Daegu High Court's 1st Criminal Division sentenced a person identified as A to two years and six months in prison for creating deepfake videos of two aespa members and selling them for profit, SM Entertainment said on the 18th. A must also complete an 80-hour sexual violence treatment programme and faces a seven-year ban on working at child- and youth-related institutions.

Graham Norton wins US order forcing Meta to name deepfake page

Who did it: HUMAN The Independent

Graham Norton won an order from the US District Court for the Northern District of California forcing Meta to reveal who runs The Westminster Wire. According to Norton, the anonymous Facebook account has posted AI-generated images of him and his husband and fabricated stories since December 2025, including false reports of his 94-year-old mother's death. He plans High Court proceedings against them in England.

San Francisco plate-reader data searched 299 times for barred agencies

Who did it: HUMAN San Francisco Chronicle

San Francisco police say a compliance audit found their Flock license plate reader data was searched 299 times over about a year by Western States Information Network analysts, via the Northern California Regional Intelligence Center, on behalf of federal and out-of-state agencies barred under California law, including the DEA, IRS and ATF. SFPD has cut off the center's access; the cameras remain on.

AI-supplied case numbers land veteran Ankara lawyer in two probes

Who did it: HYBRID Sözcü

A lawyer with 25 years in practice faces investigations by prosecutors and the Ankara Bar after a defence brief in a Kızılcahamam case cited Court of Cassation decisions with incorrect references, flagged at the hearing as fabricated. The lawyer says AI was used to add case law and got the numbers wrong but not the content, calling it an inadvertent error.

Deepfake Taoiseach promises €40,000 from a phone and €250

Who did it: HUMAN The Irish Sun

A deepfake video on social media showed Taoiseach Micheál Martin claiming anyone could earn €40,000 with "just a phone and €250", with a first €1,000 profit within hours or a personal €10,000 refund. Martin called it "clearly very false material", urged people to be vigilant, and said platforms should stop such material getting online and remove it once identified.

Waymo recalls nearly 3,900 robotaxis over closed-roadworks software flaw

Who did it: AI Reuters

Waymo is recalling nearly 3,900 robotaxis after more than a dozen incidents since early April in California and Arizona in which its vehicles failed to recognise ramp closure signs and drove into freeway construction zones, according to a filing with NHTSA. Waymo says it restricted freeway operations and updated the software; it is the company's second recall in just over a month.

Suspended year for Egyptian man's AI blackmail of relative

Who did it: HUMAN صدى البلد

A Damanhour criminal court sentenced a young man to one year with labour, suspended, and ordered the material erased, after convicting him of blackmailing a female relative with AI-fabricated pornographic images and clips sent from fake Facebook and Messenger accounts to pressure her into a sexual relationship. She and her father reconciled with him in court, prompting leniency.

Bunshun smear-video source accused of AI-made 'recovered' clips

Who did it: HUMAN ハムスター速報

According to the blog Hamusoku, citing an upcoming Shukan Bunshun article, 4 of 23 videos that Ken Matsui gave the magazine as recovered copies of smear videos he said he had made contained photos with timeline inconsistencies. Daily WiLL says Matsui most likely created them anew with AI rather than restoring them.

Sanctions motions built on AI errors earn first and final warning

Who did it: HYBRID D. Nevada

A Nevada federal magistrate judge gave defence counsel a "first and final warning" after Progressive Northern Insurance identified 11 examples of improper AI use, including fabricated or inaccurate quotations, in two sanctions motions the defendants then withdrew. The court called unverified AI content a serious breach of professional rules but denied Progressive's fee request as improperly developed.

Hallucinated citations among eight obstructions in TH Foods retaliation suit

Who did it: HYBRID D. Nevada

A Nevada federal magistrate judge ordered pro se plaintiff Preston House to show cause why case-terminating sanctions should not be recommended in his retaliation suit against TH Foods. The court lists eight kinds of obstruction, including relying on hallucinated, misquoted and miscited cases and continuing after being warned that doing so would result in sanctions.

Singapore man says ex-schoolmate faked AI family photos with him

Who did it: HUMAN The Straits Times

A Singapore man says a former schoolmate he last saw about 15 years ago posted apparently AI-manipulated images from around December 2025 presenting them as an engaged couple with a baby, some using photos of his sister's young children. He has made a police report; according to his sister, the woman said she did it out of love.