AI Misconduct.com

The daily rap sheet of the machine age.

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

Sunday, April 6, 2025

5 items

Judge sets spring 2026 jury trial in Musk v. OpenAI

Who did it: HUMAN EconoTimes

Judge Yvonne Gonzalez Rogers of the U.S. District Court for the Northern District of California set Elon Musk's lawsuit against OpenAI for a spring 2026 jury trial, denying his request to halt its for-profit transition but agreeing to expedite. Musk claims OpenAI and CEO Sam Altman strayed from the company's original nonprofit mission; they argue the suit is an attempt to undermine a key competitor.

Meta accused of training Llama 3 on pirated LibGen books

Who did it: HUMAN Natural News

Meta allegedly trained Llama 3 on millions of pirated books from shadow libraries LibGen and Anna's Archive, with court documents quoting one email: "Books are actually more important than web data." Australian authors including Hannah Kent and Sophie Cunningham, whose works appear in LibGen, are angry; Cunningham is considering legal action and asked her publishers to send Meta cease and desist notices.

Bermuda warns of AI scam video using Premier's likeness

Who did it: HUMAN The Royal Gazette

Bermuda's Cabinet Office warned of a fake AI-generated video on Facebook that featured imagery of Premier David Burt, falsely carried branding tied to The Royal Gazette and made a claim about a government investment scheme. A spokeswoman called it a scam, adding that no government official will ever use social media to promote an investment.

Xiaomi SU7 crash kills three students after assisted-driving warning

Who did it: HYBRID Gizmochina

A Xiaomi SU7 Standard crashed into a concrete guardrail on an Anhui highway on March 29, killing three university students; according to Xiaomi, its NOA assisted-driving mode was on at 116 km/h when it warned of obstacles, and the driver took control a second later before the car hit at 97 km/h. Xiaomi says AEB does not support detecting the water barrier; police are investigating.

Meta's LM Arena Maverick isn't the one developers get

Who did it: HUMAN TechCrunch

Meta's new Llama 4 Maverick model ranks second on LM Arena, but the version Meta tested there was an "experimental chat version" "optimized for conversationality" that differs from the one widely available to developers, a practice the report calls misleading. Researchers on X say the Arena version uses a lot of emojis and gives long-winded answers.