AI Misconduct.com

The daily rap sheet of the machine age.

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

Thursday, May 15, 2025

7 items

Claude invents citation details in Anthropic's own copyright defense

Who did it: HYBRID Reuters

Latham & Watkins attorney Ivana Dukanovic told a California federal judge that a footnote in an expert report for Anthropic in the music publishers' lyrics lawsuit carried a fake title and authors that Claude made up when she used it to create a citation for a legitimate journal article. She called it "an embarrassing and unintentional mistake."

FBI warns of AI voice messages impersonating senior US officials

Who did it: HUMAN ic3.gov

The FBI warned that since April 2025 malicious actors have sent texts and AI-generated voice messages claiming to come from senior US officials, targeting many current or former federal and state officials and their contacts. The bureau says the aim is to build rapport before gaining access to personal accounts, such as via a malicious link presented as a move to another messaging platform.

Refute says partly AI-made TikTok campaign targets Romanian diaspora

Who did it: HYBRID G4Media.ro

London-based Refute says it found about 32,500 TikTok videos with slogans promoting George Simion and Călin Georgescu before Romania's May 18 presidential runoff, many inauthentic, including videos duplicated across dozens of accounts, and some AI-generated. According to Refute, 48% of interactions with the videos came from outside Romania, though only 24% of Romanian citizens live abroad.

Watchdog says X apparently sold checkmarks to likely terror-linked accounts Grok flattered

Who did it: HYBRID Ars Technica

A Tech Transparency Project report says X appears to have sold paid checkmarks to accounts likely linked to US-sanctioned terrorists, some of them using X's tip buttons. The group says Grok's profile summaries gave such accounts a positive, sometimes poetic spin, calling the apparent owner of one, a sanctioned Hezbollah money exchanger, "a fierce Lebanese economist and journalist."

Soldier indicted; court papers allege AI-generated child abuse images

Who did it: HUMAN KEYE

An El Paso soldier, Seth Herrera, 35, was indicted Wednesday on charges of attempted sexual exploitation of a child and receipt of child sexual abuse material. Court documents allege he used AI chatbots to generate such material from images of children he knew; he already faces related charges in Alaska, including over AI-generated material.

Hawaii appeals court sanctions lawyer $100 over nonexistent case

Who did it: HUMAN AI Hallucination Cases database

Hawaii's Intermediate Court of Appeals sanctioned attorney James D. DiPasquale $100, payable personally, for a motion to dismiss citing "Greenspan v. Greenspan," a case the court could not find. DiPasquale said a per diem attorney drafted the motion and that, although he "did not personally use AI," he failed to verify every citation.

Korean privacy regulator faults AI textbooks' handling of student data

Who did it: HUMAN Yonhap News Agency

South Korea's Personal Information Protection Commission found the AI Digital Textbook service, running since March, fell short on personal data handling: operator KERIS omitted some required details, such as data items, purposes and retention periods, from consent forms and privacy policies. It recommended corrections and told the Education Ministry to build privacy-law requirements into textbook screening standards and guidelines.