AI Misconduct.com

The daily rap sheet of the machine age.

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

Friday, December 12, 2025

10 items

Jury acquits woman whose ex allegedly faked a text

Who did it: HUMAN The Naples Press

On December 2, after a one-day trial, a jury acquitted a woman of misdemeanor battery over scratches on her ex-boyfriend. Her defense claimed he had created a text message with a text-generating app, and that message led to her second arrest. Prosecutors had moved to exclude any mention of artificial intelligence or fabricated evidence.

Amazon appears to pull error-filled AI Fallout recap

Who did it: AI Mashable

Prime Video's AI video recap of Fallout's first season, from a feature launched in November, dated the show's flashbacks about 120 years too early and botched the main character's motivation, GamesRadar pointed out. After gaming journalists wrote about the errors, Amazon seemingly pulled the recaps from Fallout and other Prime Video series, The Verge reported.

Gambling venues use harm-reduction face scanning to boost play

Who did it: HUMAN The Saturday Paper

Facial surveillance rolled out under government initiatives to help gambling addicts and stop money launderers is being used to incentivise gambling at Australian casinos and pubs, The Saturday Paper reports. Six gambling providers' privacy policies cite secondary uses; Konami has touted AI facial recognition for "personalized" gaming, and Oosto documents say its scanning at Royal Randwick also helps "grow revenues".

Auditor general finds CRA's $18M chatbot mostly wrong

Who did it: AI National Post

The Canada Revenue Agency has spent over $18 million since 2018-2019 on Charlie the Chatbot, which Auditor General Karen Hogan's October report found answered only 2 of 6 test questions accurately, against 5 of 6 for public AI tools. The CRA says its November generative AI upgrade tested at "approximately 90%" accuracy, though it cannot precisely say how many answers are correct.

Study finds Instacart AI price tests varied identical items up to 23%

Who did it: HUMAN Fox Business

An investigation by Consumer Reports and Groundwork Collaborative found that Instacart's AI-enabled pricing experiments showed different customers prices up to 23% apart for identical items bought at the same time. Instacart did not deny the findings, calling the tests short-term and randomised, and said those on its Target storefront have ended.

Tribunal judgment quoted Forstater ruling with words it never contained

Who did it: HUMAN Scottish Legal News

A 300-page Employment Tribunal judgment by Judge Sandy Kemp quoted the 2021 Forstater ruling with words that never appeared in it, prompting suggestions that AI was used. Judge Kemp admitted the mistake and the Judicial Office called it a "clerical mistake, error or omission", but refused to explain the cause. The Telegraph reports a second fabricated quotation.

Fake AI albums appear on Doc Watson and Dillards streaming pages

Who did it: HUMAN Saving Country Music

AI-generated albums titled Roses Never Came appeared on December 10th on the official streaming accounts of Doc Watson, who died in 2012, and bluegrass group The Dillards, credited as written and performed by the artists, Saving Country Music reports. The site says the music is not theirs and likens it to a fake Don Williams album in November.

Tribunal cannot find case cited in AI-assisted skeleton argument

Who did it: HYBRID AI Hallucination Cases database

At a Reading Employment Tribunal hearing on 1 December 2025, a claimant's representative cited "Adams v Wincanton Group Ltd [2013] IRLR 725", which neither Employment Judge Anstis nor opposing counsel could find. She accepted using AI to produce the document; the judge urged checking any AI-produced case law. The interim relief claim was dismissed on separate grounds.

Río Negro court finds apparently AI-invented precedents in appeal brief

Who did it: HYBRID AI Hallucination Cases database

An appeals chamber in General Roca, Río Negro, found that four provincial high court precedents cited by a plaintiff's lawyers in a traffic-accident appeal could not be located, and inferred generative AI had been used without human control. On 12 December 2025 it rejected the appeal and, without imposing a sanction, issued a call for reflection.