A federal judge on Wednesday rejected OpenAI's request to dismiss The New York Times' lawsuit alleging the ChatGPT maker exploited the paper's content without permission or payment. Judge Sidney Stein narrowed the case, which also names Microsoft, but let its main copyright claims go forward. OpenAI says it builds its models on publicly available data under fair use; no trial date is set.
A California federal judge on Tuesday denied a preliminary bid by Universal Music Group, Concord and ABKCO to stop Anthropic using their lyrics to train Claude. The publishers sued in 2023 alleging infringement of lyrics from at least 500 songs. Judge Eumi Lee found the request too broad and saw no showing of irreparable harm; the publishers say they remain confident.
A leaked database found on an unsecured server shows 133,000 examples fed to an unnamed large language model to flag content considered sensitive by the Chinese government, from rural poverty complaints to political satire and Taiwan, TechCrunch reports. Its builder is unknown, but it says it is for 'public opinion work'; China's embassy said it opposes 'groundless attacks and slanders'.
Former NewsChannel 5 meteorologist Bree Smith told Tennessee's House Criminal Justice Subcommittee on Wednesday that, since last fall, fake Facebook accounts posing as her have used pictures putting her face on someone else's semi-nude body to try to scam her fans. Social media accounts also used her image in sexually charged AI deepfake videos. She is pushing for changes to state law.
Who did it: HUMAN Institute for Strategic Dialogue
ISD and the Bureau of Investigative Journalism identified 89 Facebook pages and 39 groups run from Sri Lanka pushing often AI-generated anti-migrant content to a combined audience of 1.6 million, apparently for profit, the researchers say. Ads included an AI image of Sadiq Khan and Keir Starmer in prison uniforms; the operator's 'online academy' showed how to use ChatGPT for content.
A deepfake video showed TODAY's Al Roker promoting a hypertension remedy and describing heart attacks; he says he has neither had a heart attack nor has hypertension. Roker said celebrity friends called because their parents were taken in. Meta said it removed the page for violating its policies after TODAY reached out; who made the video is unclear.
Scammers ran Instagram ads using an AI deepfake of a Brazilian woman with lipedema, falsely claiming a 'special tea' cleared her condition in seven days and linking to a R$37 package, BBC News Brasil reports. She reported the ad, but Instagram told her it met community standards and it was still up at publication. Meta said fraud is not allowed.
A viral 2-minute 22-second video of India's external affairs minister S Jaishankar promising ₹60,000 on day one for a ₹21,000 investment is a deepfake built from a September 2024 Asia Society address, Newschecker found. ElevenLabs' detector gave a 98% probability the voice was made with its tool, and the linked platform was rated untrustworthy.
