AI Misconduct.com

The daily rap sheet of the machine age.

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

Saturday, July 12, 2025

7 items

Gmail's AI 'translates' German emails into garbled German

Who did it: AI t-online

Gmail's AI-driven automatic translation has been mangling emails for users whose default language is not English, t-online reports: misled by English code in emails, it "translated" German newsletters into German, turning "Ukrainian positions" into "American" ones and the Israeli army into a Russian one. Google calls it a technical problem affecting a "very small number" of users and is rolling out fixes.

HKU warns law student accused of AI porn of classmates

Who did it: HUMAN South China Morning Post

The University of Hong Kong said it issued a warning letter to a male law student accused of using free online AI software to make pornographic images of classmates and other women, and demanded he apologise. Complainants claim he made more than 700 images of 20 to 30 victims, and say the university did too little to hold him accountable.

Norwegian watchdogs say Messenger AI shares data without consent

Who did it: HUMAN Nettavisen

Norway's Data Protection Authority and Consumer Council say Messenger users must switch off Meta's AI features chat by chat to avoid data sharing, which they argue means sharing without valid consent, contrary to GDPR, Digi reports. Meta itself advises against sharing passwords or financial details, warning that content with personal data may be shared with third parties.

AI video falsely shows Merz shooting polar bears in Canada

Who did it: HUMAN Зеркало недели

A fake story on a site called Toronto Journal, with an AI-generated video, claimed German Chancellor Friedrich Merz shot a polar bear, her two cubs and a seal in Coral Harbour, Canada. DW's fact-check found it to be Russian propaganda; research portal Gnida attributes it to Storm-1516, a group behind earlier fakes about Kamala Harris and Volodymyr Zelensky.

Hertz customers say AI scanner bills hundreds for tiny dings

Who did it: HYBRID Futurism

Hertz customers report steep bills from its UVeye-supplied AI damage scanner: one in Atlanta was charged $440, including $190 in fees, for an inch of wheel curb rash, and another $350 for two dime-sized bumps, The Drive reported. The app offered no human agent but discounts for paying quickly; Hertz says most rentals are incident-free.

Azerbaijan flags deepfake ministers talking about supposedly downed Belarus plane

Who did it: HUMAN Media Agentliyi

Azerbaijan's Media Development Agency said deepfake videos on social media falsely showed the country's defence and foreign ministers making statements about a Belarusian plane supposedly shot down in Russian airspace. The agency called the AI-made clips a crude example of information manipulation aimed at misleading the public and urged people to trust only official sources.