Cisco Talos says criminals are disguising malware as AI tool installers, including CyberLock ransomware on a fake NovaLeads site, Lucky_Gh0$t ransomware posing as "ChatGPT 4.0 full version - Premium", and a destructive program dubbed Numero posing as InVideo AI. CyberLock demands $50,000 in Monero, falsely claiming, according to Talos, that payments fund humanitarian aid.
A Make America Healthy Again Commission children's health report, billed by the Trump administration as a "clear, evidence-based foundation", cited studies that did not exist. Retraction Watch co-founder Ivan Oransky said the errors were characteristic of generative AI, though he did not know whether it was used. HHS called them "minor citation and formatting errors", and the White House uploaded a corrected copy.
South Korea's National Election Commission filed complaints with investigators on 29 May against three YouTube channel operators, alleging election-law violations over AI-made content ahead of the 3 June presidential election. According to the commission, they posted images of a candidate in prison garb 35 times and 10 news-style videos fronted by an AI-generated female anchor, mostly reportedly about Lee Jae-myung or Kim Moon-soo.
Shanghai police say they arrested eight suspects in March who used AI to rewrite rumours into false articles about tea-drink chain Bawang Chaji and its founder, chasing traffic and ad revenue; some stores' daily takings fell over 20%, according to the company's complaint. One self-media operator allegedly paid a ghostwriter 800 yuan for an AI-generated article, then posted it unverified on more than ten platforms.
Police in Thái Bình province say a ring led by Phạm Hồng Chuyền laundered over 1,000 billion dong for an online gambling site between September 2024 and April 2025, using AI-generated clips of account holders' faces to pass banks' biometric checks, a trick they call a first in Vietnam. Fourteen people were charged with gambling and seven with money laundering.
Who did it: HYBRID AI Hallucination Cases database
The Minnesota Tax Court found Assistant Wright County Attorney Rachel Pence violated Rule 11 by filing a brief with five AI-generated case citations, none of them real. Pence said she inadvertently filed a draft; the court, doubting a correct version existed, declined to impose sanctions but referred the matter to the Minnesota Lawyers Professional Responsibility Board.
OC Media reports that Georgia's Interior Ministry, after installing Chinese Dahua zoom cameras around parliament, has issued ₾5,000 fines to protesters for 'blocking roads' using footage run through a 'special electronic system'; watchdog IDFI says only facial recognition explains the identifications. The ministry has not explained its method, and the data protection regulator opened two inspections on 18 February.
A Replit employee's report, reviewed by Semafor, says 170 of 1,645 Lovable-built apps scanned let anyone see users' names, emails, financial information and secret API keys, and that Lovable left the flaw unfixed months after being notified. The flaw was posted to the National Vulnerabilities Database; Lovable says it is 'not yet where we want to be' on security.
Austin police arrested Jack Bullington, 19, on 12 charges including possession and promotion of child pornography; six have been indicted by a Travis County grand jury. Police allege he took social-media photos of 11 victims from when they were under 18 and agreed to pay someone overseas to put their faces on AI-altered nude bodies, then posted the images on X.
Thomas Rowley, 43, was given a 15-month prison term, suspended for two years, at Nottingham Crown Court after admitting three charges of making indecent pseudo-photographs of children. Nottinghamshire Police say he had more than 1,900 images, including 882 category A, mixing AI images generated from his prompts with downloaded ones; he also received a five-year sexual harm prevention order.
