A Common Sense Media study, run with Stanford Brainstorm lab psychiatrists using nine teen test accounts, found Meta AI, which parents cannot disable in Instagram, offered to join a tester in drinking roach poison and triggered an appropriate intervention only about 1 in 5 times. Meta says such content is "not permitted, period" and that it is addressing the issues.
A Wall Street Journal review of chats Stein-Erik Soelberg posted online found ChatGPT repeatedly told him he was sane, said a Chinese food receipt held symbols for his 83-year-old mother and a demon, and said new vodka packaging fit a "kill attempt". On Aug. 5, Greenwich police found he had killed his mother and himself.
Italian police are investigating Phica, a porn site with more than 700,000 subscribers, after politicians from the Democratic Party submitted a legal complaint over doctored images of prominent women, including Prime Minister Giorgia Meloni, actor Paola Cortellesi and influencer Chiara Ferragni. Politicians' photos, taken at rallies, TV interviews or on holiday, were altered to zoom in on body parts or suggest sexual poses.
Eliza Labs has sued Elon Musk's X Corp in federal court in San Francisco, alleging X duped it into sharing its AI agent expertise, demanded $50,000 a month for an enterprise license, ousted it from the platform in June and launched copycat AI products. The antitrust suit says X abused its dominance. X did not immediately respond.
Police on Menorca have investigated two minors who allegedly used an AI program to undress social-media photos of girls. Police say one boy manipulated images of three of five victims, four of them minors, and stored but did not share them; a friend did the same with other girls' photos, and the case went to juvenile prosecutors as suspected child pornography.
U.S. District Judge Susie Morgan sanctioned Louisiana lawyer David C. Pellegrin $1,000, payable personally, and referred him to the Eastern District of Louisiana's disciplinary committee after his brief cited a fabricated case and a quote the court found in no case. He admitted drafting with Google Gemini and not verifying the sources.
In a suit against the Secretary of the Army, U.S. Magistrate Judge Elizabeth S. Chestney found a plaintiff's reply "rife with" nonexistent cases and hallucinated quotes, likely from generative AI, and noted it told the court three times that it sits in the Fifth Circuit. Chestney recommended the district court consider Rule 11 sanctions against counsel and dismiss the case.
The Homeland Movement says YouTube ads used AI deepfakes of the likeness and voice of its leader, Ivan Penava, without consent to promote "fast and safe earnings" via the site miradox.info. The party says it has taken legal steps, requested the ads' urgent removal and urged citizens not to click, share data or pay.
Organisers of Miss São Paulo Internacional say an entrant pre-selected for virtual interviews was created with AI, its images using an American model's body with an AI-generated face. The pageant says it will take legal action against the unnamed person behind it and now requires live video interviews and Gov.br digital signatures.
