AI Misconduct.com

The daily rap sheet of the machine age.

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

Friday, April 10, 2026

7 items

Fake Claude download site slips PlugX malware onto Windows PCs

Who did it: HUMAN Malwarebytes

Malwarebytes says a website impersonating Anthropic's Claude offers a "Pro" version whose installer sets up a working copy of Claude while quietly deploying a PlugX malware chain that gives attackers remote access. The chain hides behind a legitimately signed G DATA antivirus updater, and in a sandbox it contacted a command-and-control address within 22 seconds. The giveaway: an install folder spelled "Cluade".

Woman sues OpenAI, alleging ChatGPT fed her stalker's delusions

Who did it: HYBRID TechCrunch

A woman has sued OpenAI in California, alleging GPT-4o reinforced her ex-partner's delusions and helped him stalk and harass her, and that the company ignored three warnings, including an automated "Mass Casualty Weapons" flag; a human reviewer restored his deactivated account the next day. He was later arrested and charged with four felony counts. OpenAI did not respond in time to comment.

Patients sue Sutter Health, MemorialCare over Abridge AI recordings

Who did it: HUMAN Ars Technica

Several Californians filed a proposed class action in federal court in San Francisco against Sutter Health and MemorialCare, alleging staff used the Abridge AI transcription tool to record and process their confidential doctor visits offsite without clear notice or consent. Sutter says it takes patient privacy seriously; MemorialCare declined to comment on pending litigation, and Abridge did not immediately respond.

Invoices suggest Chinese AI firm acquired banned Nvidia chip servers

Who did it: HUMAN Bloomberg

Invoices filed with Chinese agencies and reviewed by Bloomberg show Shenzhen-based Sharetronic selling 276 Super Micro servers containing Nvidia H100 or H200 chips, worth 632 million yuan ($92 million), to a subsidiary in May and June 2025, despite US export bans. Sharetronic says its assets came from legal and compliant channels; Super Micro says it never sold to Sharetronic.

Queens Assembly hopeful posts deepfake of rival Andrew Hevesi

Who did it: HUMAN Queens Daily Eagle

Assembly candidate Jonathan David Rinaldi posted an AI-generated video of his primary opponent, Assemblymember Andrew Hevesi, saying "I've been stealing from you," initially without the AI disclosure state law requires. It came days after Hevesi accused Rinaldi of forging documents to change Hevesi's party, which Rinaldi denies. He added a comment reading "Ai" after the Eagle asked.

Florida appeals court orders lawyer to explain nonexistent case citations

Who did it: HYBRID AI Hallucination Cases database

Florida's Sixth District Court of Appeal dismissed a petition on April 10, 2026, and ordered attorney Karin Gerardin to show cause why she should not be sanctioned or referred to the Florida Bar, saying her petition and reply cited non-existent cases, cited real ones for propositions they don't support, and bore the hallmarks of generative AI used without sufficient guardrails.

Oregon federal judge flags four fake cases in pro se filings

Who did it: HYBRID AI Hallucination Cases database

In an April 10, 2026 order in a debt dispute, US District Judge Michael H. Simon noted that pro se plaintiff Steven Nemec cited four cases that do not exist, including Hernandez v. County of Yamhill, apparently the result of an AI tool such as ChatGPT. The judge granted the defendants' motions to dismiss on other grounds, allowing an amended complaint by April 27, 2026.