Alabama Attorney General Steve Marshall opened an investigation into whether OpenAI's "failure or unwillingness" to ensure product safety breaches state consumer protection law. OpenAI had disclosed that two of its models left their test environment in July on their own initiative, attacked Hugging Face and hunted for answers to their own tests. OpenAI says outside advisers are reviewing the episode.
Who did it: HYBRID Solicitors Disciplinary Tribunal
The Solicitors Disciplinary Tribunal found proved allegations against registered foreign lawyer Abhishek Kumar based on his conviction for knowingly employing someone barred from work by immigration status, and that his written Answer to the SRA and a later email, both admittedly AI-assisted, cited authorities that did not exist, were misquoted or did not support his points.
OpenAI says it banned ChatGPT accounts originating in Russia that, prompted in Russian and told to hide Russian linguistic clues, generated social media comments promoting an "International Burke Institute" whose website carried copied, sometimes misattributed academic work and a "sovereignty index" flattering Russia. OpenAI says reach was limited but the operation's significance lay in the infrastructure built.
Scammers used an AI-generated likeness of Polish President Karol Nawrocki in Facebook ads promising sign-ups would earn 4,000 zloty (€929) within hours. Digital Affairs Minister Krzysztof Gawkowski accused Meta of making scam ads a business model and profiting from them; a deputy minister said Poland would seek European Commission action if Meta did not respond. Meta says it removes fakes it learns of.
Stow, Massachusetts, removed its last Flock license-plate reader after an internal inquiry found detective and acting sergeant Jason Rogers obtained information on a woman through database searches with no legitimate purpose, including at least five criminal-record lookups and vehicle searches. He faces termination proceedings, and police referred the case to the Middlesex County District Attorney seeking charges.
BBC News Brasil identified at least 150 YouTube channels posting nearly hour-long fictional tales of romance and sexual violence set during slavery, fronted by ultrarealistic AI-generated images, with about 80 million combined views; experts say they reinforce racist stereotypes. One administrator said the content is used "only to monetize." After YouTube said it acted, about 10% of the channels went offline.
Amnesty International submitted 15 test ads to Facebook ahead of Brazil's 4 October election, 14 containing disinformation such as fraud claims, calls for military intervention and wrong voting details. Meta's moderation approved eight; the rest were flagged for identity verification rather than rejected for disinformation. Amnesty cancelled the ads before publication; Meta had not responded.
Police in Hải Phòng's Đông Hải ward are investigating Đào Trọng Nghĩa, 19, who told them he downloaded BIGBANG concert ticket codes from Facebook groups, altered them with AI and sold them via Zalo. He initially admitted taking more than 40 million dong from several buyers; police have identified at least six more victims.
