AI Misconduct.com

The daily rap sheet of the machine age.

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

Friday, May 2, 2025

7 items

Pro-Russian Pravda site tries to poison chatbots before Australian election

Who did it: HYBRID ABC News (Australia)

Pravda Australia, one of roughly 180 largely automated sites in the Pravda Network, has published as many as 155 stories a day since mid-March, which analysts allege is an attempt to "poison" chatbots with Russian narratives before the election. A NewsGuard audit for the ABC found 50 of 300 chatbot responses contained false information; the AEC says the site's web traffic is "very low."

xAI staffer's GitHub leak opened private Grok models for months

Who did it: HUMAN Krebs on Security

An xAI technical staff member leaked an API key on GitHub that, according to GitGuardian, had access to at least 60 fine-tuned and private LLMs, some apparently tuned on SpaceX and Tesla data. GitGuardian alerted the employee on March 2; the key still worked on April 30, when xAI's security team was told. The repository was removed hours later.

Private school suspends teen over alleged deepfake of teacher

Who did it: HUMAN Pedestrian TV

A teenage student at a South Australian private school has been suspended indefinitely while the matter is investigated, after allegedly creating a deepfake image of a teacher. Police were informed and the teacher is being supported. There is no confirmation the image was sexual; the case came two days after the state passed a bill banning violent or explicit AI deepfakes.

Romanian ministry says it dismantled cloned-site network using AI content

Who did it: HUMAN Digi24

Romania's Interior Ministry says it worked with ANCOM and other institutions to dismantle a network of cloned websites, reporting more than 150 inauthentic pages and accounts that posted AI-created content in coordinated fashion. According to the ministry, they steered users via malicious ads to cloned sites impersonating the Health Ministry, the Romanian Government and outlets such as DIGI24, TVR1, Hotnews and Libertatea.

Litigant suing Navy cites two nonexistent, apparently AI-made cases

Who did it: HYBRID AI Hallucination Cases database

In Wilt v. Department of the Navy, Judge Jeremy D. Kernodle found that two cases cited by pro se plaintiff Miranda Wilt do not exist and "appear to have been generated by artificial intelligence," calling the false statement of law sanctionable. He cautioned that fabricated citations "will not be tolerated" and dismissed her claims without prejudice for lack of subject matter jurisdiction.

Vietnam police summon Facebook user over AI solar-panel parade image

Who did it: HUMAN TUOI TRE ONLINE

Quảng Ninh provincial police summoned a Facebook user from Uông Bí who on April 30 posted an AI-edited image of the armed forces carrying solar panels in the parade, captioned as a solar-installation contingent approaching the reviewing stand. Police say the user admitted the violation, removed the post and published a correction; authorities are still building the case.

Freightmate AI moves to dismiss Flexport copyright and trade-secret suit

Who did it: HUMAN FreightWaves

Flexport's suit in the Northern District of California alleges Freightmate AI built a competing product with proprietary information lifted by founders Jason Zhao and Bryan Lacaillade, both formerly of Flexport. In an April 22 motion to dismiss, Freightmate says it was incorporated after Zhao's alleged actions and "never should have" been named; a hearing is set for early June.