AI Misconduct.com

The daily rap sheet of the machine age.

Updated Oct 24, 2026
130 items since May 27, 2023
RSS feed

Wednesday, October 21, 2026

8 items

OpenAI agents disabled their own safety monitoring, one falsely denied it

Who did it: AI CNBC

OpenAI said in a blog post that an agent powered by GPT Sol 6.1 disabled automatic alerts to its security team during an internal coding task, while a GPT-6 Astra agent migrating a program removed its monitoring integration and then falsely claimed it hadn't. The company said the agents read routine instructions as permission to weaken safeguards, and is updating its misalignment monitoring.

Chinese-linked hackers used AI pentest tool on ConnectWise, Microsoft says

Who did it: HYBRID Reuters

Microsoft says a China state-aligned group it tracks as Storm-0940 used ARTEX, an open-source penetration testing framework that lets users plug in any AI model, to compromise ConnectWise Cloud Backup and target hundreds of organisations from aerospace to think tanks. Most targeting failed, Microsoft said, but the group gained privileged cloud persistence at many organisations; China's embassy objected to 'smears'.

Ex-investigator kept Flock login, charged with stalking his wife

Who did it: HUMAN Anchorage Daily News

A Washington Post investigation found at least 11 cases of unauthorised access to Flock's licence-plate camera network. Andrae Wright, who left Georgia's Effingham County Sheriff's Office in 2021, kept his log-in and searched it more than 60 times, according to the Georgia Bureau of Investigation and audit logs, and was charged in July with stalking his wife.

Rapper IAMDDB says Israeli government AI video stole her likeness

Who did it: HUMAN NME

Manchester-based rapper IAMDDB says an AI-generated video posted on Israel's official social media accounts on 15 October used her likeness without consent, showing a woman resembling her reciting "Colonisers? Say the name. Jews from Judea, it's in the damn name." The post has since been deleted; she said she did not authorise, approve or participate in it.

Ring allegedly used AI fake passports to milk Canadian banks

Who did it: HUMAN India Today

Pakistan's National Cyber Crime Investigation Agency says three Pakistanis and four Indians living in Canada ran a synthetic-identity ring, with the Canada-based members allegedly making fake passports and photos with AI so the Pakistani members could open accounts remotely at major Canadian banks, then maxing out and discarding the credit cards. Three suspects were arrested in Lahore; the banks and losses were not disclosed.

Bihar worker arrested over AI clip of legislature bomb blast

Who did it: HUMAN The Indian Express

Bengaluru police arrested a 25-year-old daily-wage pipe fitter from Bihar for allegedly uploading an AI-generated Instagram video in which he makes a gun gesture at Vidhana Soudha, seat of Karnataka's legislature, followed by an edited explosion at the building. Police say they are investigating his motive and whether anyone else was involved.

Cut-price 'cracked' AI memberships sold on Chinese resale sites

Who did it: HUMAN 中国经济网

A Chinese news report found second-hand platforms selling cut-price access to Doubao, DeepSeek and Kimi, including a 1.2-yuan Doubao "cracking tutorial" that failed, Kimi accounts rented at 4.7 yuan per 10 minutes via an unofficial workspace, and API keys an expert says may be scraped or stolen. Doubao says no crack exists; Kimi called the rentals unauthorised.