Publishers owning nearly 400 newspapers sued OpenAI and Microsoft in the Southern District of New York. The complaint alleges the companies "systematically and secretly crawled" their sites to train ChatGPT and Copilot, stripped copyright management information and reproduced articles, and that none of the billions in resulting value reached the publishers. OpenAI says its models are grounded in fair use.
A proposed federal class action filed Monday accuses Marathon, Circle K, BP, Walmart, Albertsons and other California fuel retailers of using AI-powered Kalibrate software to collude on pump prices. The suit calls the software the "central nervous system" of a conspiracy and says a single cent at the pump costs California drivers $134 million a year.
In ruling 23006, the Third Criminal Section of Italy's Court of Cassation increased the pecuniary sanction on a defence lawyer whose appeal cited non-existent precedents produced by "computer hallucination". The court said using AI does not excuse the duty to verify sources. It held that citing invented case law shows aggravated fault, not excusable error.
Who did it: HUMAN OFFICE OF THE DISTRICT ATTORNEY QUEENS COUNTY
Jonathan Rinaldi, a 2025 candidate for New York City Council District 29, was charged with forgery for allegedly using AI to fabricate endorsements and news stories. These included a fake New York Post article with a face-swapped photo showing him with former Council Member Robert Holden. According to prosecutors, he defended an altered endorsement as "asymmetrical warfare." He faces up to two years if convicted.
Police are investigating two pupils at Berlin's Heinz-Brandt-Schule on suspicion of creating and spreading AI-generated sexualised nude images of female classmates, reportedly in at least three cases. The pair were suspended. Students who allegedly knew but did not report also face measures. The state criminal police unit for sexual offences against minors has taken over the case.
A Munich actress told a press conference that she once found a realistic pornographic deepfake bearing her face through a search engine. The image vanished after a few months, and she never went to police. Bavarian Justice Minister Georg Eisenreich, who hosted her, wants up to five years in prison for serious cases and called a federal draft's two-year maximum too low.
SentinelLabs says a North Korea-linked Rust backdoor for macOS, dubbed macOS.Gaslight, carried 38 fabricated system messages meant to push AI malware-triage tools into aborting analysis. Behind the injection sat a stealer that targets browser data, terminal histories and the login keychain. It is controlled through Telegram's Bot API. The researchers advise treating sample contents as adversarial input.
Between March 29 and April 6, a 28-year-old man sent threats to kill, set fires and plant bombs to Taipei Metro's AI customer-service chatbot five times. The Taipei District Court sentenced him to six months, commutable to a fine. He confessed, and the sentence can be appealed. Only staff saw the messages.
Rep. Anna Paulina Luna (R-FL) said her staff used AI only to spell- and grammar-check a 2027 NDAA amendment summary. She said this after X users circulated a version containing the words "Claude responded:". Her first post said staff "used AI to correct a draft text and didn't edit." She later insisted "NO Legislation is ever drafted with AI."
The Court of Appeals of Georgia found that a pro se appellant's two-sentence argument in a divorce fee dispute cited just one case, which does not exist. The court still vacated the trial court's attorney fee award against her and remanded for a new order.
