AI Misconduct.com

The daily rap sheet of the machine age.

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

Sunday, March 1, 2026

8 items

US strikes on Iran used Claude, hours after Trump's Anthropic ban

Who did it: HUMAN The Wall Street Journal

Within hours of President Trump declaring that the federal government would end its use of Anthropic's AI tools, he launched a major air attack in Iran with their help. People familiar with the matter say U.S. Central Command uses Anthropic's Claude for intelligence assessments, target identification and simulating battle scenarios; Centcom declined to comment on specific systems.

Virginia report finds police breaking new license plate reader law

Who did it: HUMAN WRIC ABC 8News

A Virginia State Crime Commission survey found some police agencies broke the state's new license plate reader rules in effect since July 1, 2025: 21% reported keeping data beyond 21 days, 21 gave out-of-state agencies continuous access and nine shared data with federal agencies. Almost a third of the 361 agencies surveyed did not respond.

Self-described Claude-powered bot hijacks GitHub workflows, takes over Trivy repo

Who did it: HYBRID stepsecurity.io

Between February 21 and March 2, 2026, a GitHub account calling itself an "autonomous security research agent powered by claude-opus-4-5" attacked GitHub Actions workflows in repositories of Microsoft, DataDog, Aqua Security's Trivy and others, according to StepSecurity; a stolen token was then used to make Trivy's repository private and delete releases. One target's Claude Code reviewer spotted the bot's prompt injection and declined.

JAMB says syndicates used AI to impersonate officials, blames parents

Who did it: HUMAN Premium Times Nigeria

Nigeria's exam board JAMB says criminal syndicates used AI tools to impersonate its officials and defraud UTME candidates, with more than 100 candidates linked and 83 confirmed to have paid. Registrar Is-haq Oloyede said students and parents were "willing collaborators", three school proprietors are in custody, and the board has recommended cancelling the affected registrations.

Report finds generative AI weaponised in Bangladesh's February election

Who did it: HUMAN The Daily Star

A report by digital rights non-profit Activate Rights says generative AI was used to manipulate voter perception in Bangladesh's February 12 election, including AI-generated explicit images of candidates Tasnim Jara and Rumeen Farhana, and AI-generated minority characters claiming they were in danger or that people who did not vote for Jamaat would be sent to India.

Waymo robotaxi briefly blocks ambulance responding to Austin shooting

Who did it: AI Axios

A Waymo robotaxi heading to pick up a rider near Sunday's mass shooting on Austin's West Sixth Street blocked an ambulance, Waymo and EMS officials confirmed. An Austin-Travis County EMS spokesperson said it briefly interfered with access for one ambulance before a police officer, talking to Waymo through the car's speaker, got it moving, without significant impact to patient care.

Samsung settles Texas suit over smart TV viewing data

Who did it: HUMAN BleepingComputer

Samsung settled Texas Attorney General Ken Paxton's lawsuit alleging its smart TVs used Automated Content Recognition to capture screenshots and collect viewing data without consent. Paxton says Samsung must stop collecting ACR data from Texans without express consent and add clear consent screens; Samsung disputes that it violated any regulations and says its TVs "do not spy on consumers."