Manhattan prosecutors seized the domains of 12 websites selling AI-generated deepfake pornography of about 1,200 real people, overwhelmingly women and mostly public figures, District Attorney Alvin Bragg said. The sites, with names like AI Celebs and Deepfake Porn, were seized after one survivor's complaint; no one has been charged, and Bragg conceded many more remain.
Spain's data protection agency has received its first breach notification in which the attack was reportedly carried out by an AI agent using a well-known language model, which logged in, autonomously found application vulnerabilities, then modified personal data and accessed invoices, writes Francisco Pérez Bes. The article stresses the account comes from the affected organisation and awaits analysis.
A Chinese student, then 23, who admitted stabbing three five-year-olds in Zurich-Oerlikon on 1 October 2024 used ChatGPT to discuss the legal consequences of random stabbings and, with Google, to research where and when to find children, according to the indictment. The Zurich District Court trial covers several attempted-murder counts; prosecutors seek 15 years and a 15-year expulsion.
OpenAI is paying hundreds of contractors to read and rate real ChatGPT conversations under codename Project Lily, 404 Media reported; reviewers see no usernames but sometimes see memory summaries hinting at past use and location. OpenAI says a Privacy Filter model strips personal data first, though its own page says it can miss identifiers; it named a disclosure help page only after publication.
Judge Seibel of the Southern District of New York ordered plaintiff's lawyer Brian Higbie to show cause why his opposition briefs in a suit against the Village of Mamaroneck, which quoted passages that do not exist and misstated holdings, did not violate Rule 11. The errors went unaddressed for months after opposing counsel flagged them; his response is now before the court.
Who did it: HUMAN Federal Circuit and Family Court
Australia's Federal Circuit and Family Court ordered costs personally against Melbourne sole practitioner Gurdial Singh Ranjit Singh, who failed to disclose drafting a migration applicant's filings, acting 'in the shadows', and whose submissions cited AI-hallucinated authorities or cases that did not support the propositions cited. Judge Bingham found breaches of duty to the court and considered sanctions for his AI misuse.
After the military recovered a drone from the sea near Jarosławiec on Monday, an AI-generated photo of a wreck marked '540' circulated online and was unknowingly used by some media, prompting a warning from the Prime Minister's chancellery. A detection tool rated it 83.5% likely AI-made, and the Operational Command spokesman confirmed it was fake.
Fourteen parties backing Denis Bećirović, Chairman of Bosnia and Herzegovina's Presidency, accused Bakir Izetbegović's SDA of orchestrating a smear campaign featuring AI-generated videos that put words in Bećirović's mouth, along with doctored interviews and fake articles, ahead of the 4 October 2026 vote. They predict the SDA and its 'mercenaries' will keep spreading fakes.
A 29-year-old Bengaluru security guard lost over Rs 1.04 lakh after a WhatsApp video caller, allegedly a deepfake of Tamil Nadu chief minister C Joseph Vijay, offered financial aid on 9 September, and a 'manager' promising Rs 11 lakh and a supposed CBI officer then demanded fees and fines, according to the complaint. Police registered a case under the Information Technology Act.
Roughly 1,000 pages obtained by the Electronic Frontier Foundation through a FOIA lawsuit show CMS's WISeR pilot, where vendors use AI to vet Medicare prior authorizations in six states, sometimes missed its 72-hour turnaround, with one request unanswered for 83 days. Ohio vendor Innovaccer had warned CMS before launch that its system was insufficiently tested; CMS declined to delay.
Florida's First District Court of Appeal rejected a pro se appeal against DeepGulf, Inc. and Toke Oil and Gas and warned the appellant, whose briefs were 'replete with' citations to nonexistent or unrelated cases such as 'Jackson v. Household Fin. Corp.', that future fabricated sources may bring sanctions. The court never mentions AI.
