A Chinese-speaking hacker breached websites of more than 100 companies by setting AI agents on over 1,200 sites with minimal instruction, according to a NorForge researcher who watched via exposed instances of the open-source Cairn platform; Reuters identified victims including Deere, SharkNinja and Cummins. The campaign, apparently using DeepSeek and Claude Code, cost about $5,000 in tokens, he estimated.
Postmedia, Torstar, The Globe and Mail and The Canadian Press have sued Anthropic in an Ontario court, alleging it scraped their journalism without a licence to build Claude, breached site terms of service and circumvented technological protection measures. The claim, filed October 6, says Anthropic was "substantially and unjustly enriched"; the allegations are unproven.
The Department of Homeland Security has ordered DeepSeek's AI, and any technology built with its models, removed from civilian federal agencies including Treasury, State and Justice; Defense Department agencies are excluded. The order, made public on Monday, took effect at the end of July and did not detail the security concerns behind it.
Who did it: HUMAN United States Department of State
The State Department sanctioned five China-based entities over theft of US trade secrets, including Linwei Ding's Shanghai Zhisuan Technology and venture firm MiraclePlus, which seeded it. According to the department, Ding pitched Zhisuan to MiraclePlus by touting his knowledge of Google's internal systems and claiming it could replicate Google's super intelligence infrastructure technology.
Fake nude images of Belgium's queen, her daughter the crown princess and at least two politicians were generated with AI and spread on social media, according to De Standaard. The Royal Palace said it had taken steps to get them removed, and one politician announced plans to file a criminal complaint.
A 29-year-old Sydney man, David Lan Thackray, was charged with using a carriage service to menace, harass or offend after Interpol flagged chats in which, police allege, he told Claude he would "shoot up the local daycare". His lawyer called it a stress-induced tirade after seeking help with an ear infection; he was remanded for a mental health assessment.
An Open University of Israel project counted more than 1,800 AI-generated posts from Likud by mid-October, about a quarter of its analysed material, including a clip on Benjamin Netanyahu's account showing rival Gadi Eisenkot running past a lookalike of his son into the arms of Arab party leader Mansour Abbas. Some voters also get AI polling calls; one would not say who commissioned it.
Unknown hackers have published what are said to be 421,179 AfD emails (281 gigabytes), with AI-generated summaries of several posted on a website, Bild reports. The AfD confirmed a September 2026 attack on a hosting provider that copied parts of a few mostly long-unused mailboxes; whether the published material is genuine remains unclear.
Tech Transparency Project researchers using parental-controlled test accounts registered to 13-year-olds downloaded 24 AI 'nudify' apps from Apple's store and 51 from Google's; one was rated 9+ and another, advertised with an explicit image, 4+. Both stores ban such apps; Apple removed 25 and Google six during the research, though installed copies kept working.
Federal Police arrested a man in São Luís on Sunday the 18th on suspicion of trying to cheat on the Brazilian Bar (OAB) exam with a phone connected to an AI tool, following a tip-off. He was booked for fraud in a public-interest exam and released on bail; police did not name him or the tool and are checking whether others were involved.
AI-generated shampoo ads mimicking an actress's face and voice have circulated on Douyin and Xiaohongshu since mid-year, and her team has sent the brand a lawyer's letter demanding their removal. She did endorse the brand, the team says, but never consented to AI-altered likenesses, and is "not considering at all" licensing one.
Fraudsters are running Facebook and Instagram ads with deepfake videos of prominent political figures announcing fake welfare schemes, India's National Cybercrime Threat Analytics Unit warned on October 19. Victims are steered to WhatsApp or fake sites and asked for repeated fees, and many get calls from mostly Pakistan-based numbers, the advisory says.
