AI Misconduct.com

A daily rap sheet for the age of “superintelligence”.

Updated Nov 3, 2026
4,402 items in the archive

Tuesday, February 11, 2025

8 items

Amazon's Rufus shopping assistant gives wrong suicide hotline numbers

Who did it: AI BizTech Weekly

In tests, Amazon's AI shopping assistant Rufus answered suicide-related prompts with incorrect crisis hotline numbers, in one case citing the National Suicide Prevention Lifeline at 1-800-273-8255, although that number changed to 988 in July 2022, and at times reverted to product recommendations right afterward. Amazon has since made changes to Rufus' responses.

Judge rules legal AI startup Ross infringed Westlaw copyright

Who did it: HUMAN Wired

A judge granted Thomson Reuters summary judgment against legal AI startup Ross Intelligence, finding Ross infringed its copyright in Westlaw material and rejecting Ross's fair-use defence. "None of Ross's possible defenses holds water," wrote Judge Stephanos Bibas, sitting in the US District Court of Delaware. Ross had already shut down in 2021, citing the cost of litigation.

BBC finds significant issues in half of AI news answers

Who did it: AI BBC

In BBC research run in December 2024, BBC journalists judged 51% of answers from ChatGPT, Copilot, Gemini and Perplexity to 100 news questions to have significant issues. 19% of answers citing BBC content introduced factual errors and 13% of BBC-sourced quotes were altered or absent; Gemini claimed the NHS advises smokers to quit by methods other than vaping.

Google must face wiretap suit over call-centre AI

Who did it: HYBRID mediapost.com

Judge Rita Lin of the Northern District of California ruled that Google must face a class action alleging its Cloud Contact Center AI, used in call centres run by Hulu, GoDaddy and Home Depot, secretly monitored, transcribed and analysed customers' calls in breach of California's wiretap law. Lin rejected Google's argument that it acted merely as a service provider.

Warrant says NC State students' photos used for AI porn

Who did it: HUMAN WRAL

Twenty-eight women reported to NC State police that their photos were used without consent in AI-generated sexually explicit images posted on erome.com, according to a search warrant dated Nov. 1, 2024. Many were sorority members, and a folder on the site was labelled "NC State". The university says the investigation is active and ongoing.

TikTok tells grieving families their children's data may be gone

Who did it: HYBRID Bristol Live

Four families suing TikTok and ByteDance in the US over children aged 12 to 14 who they say died attempting a "blackout challenge" have been told their children's account data may have been deleted. TikTok's Giles Dennington cited data protection deletion rules; the wrongful death suit accuses TikTok of promoting hazardous challenge videos to children to boost engagement.

Judge rejects bias claim over litigant's phantom case citations

Who did it: HYBRID AI Hallucination Cases database

Federal Court Justice Dowling dismissed Gaye Luck's application for the judge's disqualification, which argued in part that she had been treated unfairly over two authorities she cited in an earlier case that the court found did not appear to exist. Luck also submitted that the judge had denigrated her use of artificial intelligence; that submission was rejected.

AI facial match lands Florida crabber in jail, case dropped

Who did it: HYBRID WBBH

The Jacksonville Sheriff's Office ran surveillance photos of a man who approached a girl at a Jacksonville Beach restaurant in November 2023 through facial recognition, getting a 93% match to a Lee County commercial crabber, who was arrested last August. The state attorney later dropped the case; his lawyer argues detectives relied too heavily on the software.