According to Push Security, attackers are promoting shared pages on chatgpt.com and claude.ai through search ads to deliver malware. In a new variant, ChatGPT's code rendering is used to fake a service-disruption notice. The notice sends users to a cloned download page serving a malicious 'ChatGPT for Desktop' executable, while Claude chats posing as install guides push curl commands.
Ottawa police have charged two men after an investigation into AI-generated violent and sexual images of women made from their social media photos. Court documents identify up to 25 alleged victims. Stephen Lowe, 60, of Maitland, N.S., faces 79 charges, including child sexual abuse material counts, and Gregry Peter Van Beek, 38, faces seven. The allegations are unproven.
Meta's Model Capability Initiative records US employees' clicks, mouse movements and navigation across more than 200 apps and websites to train AI agents. According to internal documents seen by Reuters, it also captures emails and chats sent by colleagues outside the US. Privacy group NOYB says this could breach GDPR, and Meta says it mitigated the privacy risks.
Who did it: HUMAN Texas Department of Public Safety (.gov)
Texas DPS says Adan Covarrubias, 31, of Carrizo Springs, has been charged with possession of computer-generated child pornography, a first-degree felony, under the state's new Senate Bill 20. Investigators say he made more than 900 AI images and videos by 'nudifying' children's photos taken from school-affiliated pages. About 30 child victims have been identified so far.
Democratic Party candidate Kim Kyoung-soo's camp filed a police complaint against five people. Citing a former staffer's account to JTBC, it alleges that People Power Party rival Park Wan-su's camp made AI deepfake videos attacking Kim, with material from a provincial government official. Park's camp denies this, and the provincial election commission says it is investigating.
Security firm WithSecure says GREYVIBE, a Russian-linked group targeting Ukrainian military, government, civilian and business organisations since at least August 2025, used ChatGPT and Google Gemini for code and Ideogram AI for images. The AI help backfired: flaws it introduced into the group's LegionRelay trojan exposed backend functions, and the group also uploaded test samples to VirusTotal.
The Delhi High Court issued summons to unnamed defendants and said it would pass an interim order protecting Telugu actor Naga Chaitanya's personality rights. His suit alleges misuse of his identity through pornographic websites, AI-generated content, deepfake videos and unauthorised merchandise, and seeks a dynamic injunction. The matter returns on September 30.
Ahmedabad Cyber Crime Branch police say they arrested Sumit Sharma, 27, of Delhi. They allege he searched for tools such as 'AI Remove Clothes' to make morphed nude images and videos of a woman and her mother after she refused his advances. Police found more than 100 such images and eight to 10 fake accounts in her name.
Who did it: HYBRID AI Hallucination Cases database
Paraná's Court of Justice fined a defendant 1% of the updated case value for bad-faith litigation. The defendant's appeal had cited non-existent precedents, including a supposed STJ ruling and decisions attributed to the Paraná court itself. The 19th Civil Chamber said these were possibly generated by AI tools without checking. The appeal itself was dismissed for unpaid court costs.
A Tesla owner who paid $10,000 for Full Self-Driving in 2021 won a $10,672.88 default judgment in Travis County, Texas, small claims court on April 1, arguing the promised self-driving was never delivered. Tesla, which maintains it did not breach the contract, lost a May 13 rehearing over its missed deadline; the owner received a $10,800 check May 29.
