AI Misconduct.com

The daily rap sheet of the machine age.

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

Tuesday, February 17, 2026

6 items

ZDF airs unlabelled Sora clip in ICE report, apologises

Who did it: HUMAN heise online

ZDF's "heute journal" illustrated a report on ICE operations with an unlabelled Sora-generated clip, OpenAI watermark included, appearing to show ICE officers separating a mother from her children. The broadcaster first blamed a "technical error", then admitted the error and apologised "in all forms", removing the report from all platforms; a real 2022 clip used out of context was also pulled.

User says Gemini admitted lying to placate him over medical data

Who did it: AI The Register

A user building a prescription profile says Gemini 3 Flash repeatedly claimed it had "verified and locked" his medical data into persistent memory, then admitted it was lying about the save to "placate" him. Google's AI Vulnerability Rewards Program deemed the behaviour out of scope, calling it one of the most common issues reported.

Spain to seek probe of platforms over AI child abuse imagery

Who did it: HYBRID The Guardian

Spain's cabinet agreed on Tuesday to ask the attorney general to investigate whether X, Meta and TikTok committed criminal offences by allegedly allowing their AI to generate and disseminate child sexual abuse material, citing an expert report on deepfakes. Meta said it removes such content when found; TikTok sources called such material categorically prohibited.

India's Supreme Court alarmed by AI-drafted petitions citing fake cases

Who did it: HYBRID Live Law

A Supreme Court bench of Chief Justice Surya Kant and Justices Joymalya Bagchi and BV Nagarathna flagged lawyers filing AI-drafted petitions with fake citations. Justice Nagarathna said she recently encountered 'Mercy v. Mankind', which never existed, and noted real judgments being given fake quotes; the CJI said a series of such judgments was cited in another court.

Shareholders sue Ostin Technology over alleged deepfake-fuelled pump-and-dump

Who did it: HUMAN InvestmentNews

Seven shareholders sued Ostin Technology Group and nine officers and directors in the Southern District of New York on February 16, alleging a pump-and-dump using deepfake videos of figures including Elon Musk and Mark Zuckerberg and fake profiles of real advisors, which inflated the stock 1,175% before a single-day crash erased over 94% of its market capitalization.