Texas sues OpenAI, and Elvis's rights holders sue Anthropic

Texas Attorney General Ken Paxton sued OpenAI and its chief executive Sam Altman under the state's Deceptive Trade Practices Act, saying the company marketed ChatGPT as safe while it aided mass shooters, encouraged suicide and left minors "addicted to a tool that fakes human compassion." According to KBTX, the suit was filed in Collin County and asks for an order blocking minors from ChatGPT and requiring a court-appointed evaluator to approve new models before release. Texas follows Florida, which sued in June, and the eight states that filed on October 15. OpenAI did not immediately respond.

Authentic Brands Group, which controls Presley's music and likeness, sued Anthropic on Friday, Billboard first reported, alleging Claude was trained on 274 of his songs, from "All Shook Up" to "Don't Be Cruel." The suit seeks the statutory maximum of $150,000 per song, roughly 41 million dollars in all. It says that even after Anthropic added guardrails, Claude recreated songs "for a modern audience" and conceded the results were "edging past 'inspired by' into reproducing the copyrighted song." It follows the August suit from Sony Music Publishing and Warner Chappell, and is one of the first brought by an artist's estate rather than a label.

CNBC's All-America Economic Survey of 1,000 registered voters, fielded October 14 to 18, found that 63 percent oppose an AI data center within 10 miles of their home and only 13 percent support one. The partisan split is the news: 55 percent of MAGA supporters want companies to keep developing advanced AI while Congress writes rules, yet just 29 percent would accept a data center nearby. Among Democrats, 80 percent oppose one and 4 percent support one. The proximity question went to about 500 voters, so its margin of error is roughly 4.4 points. It follows Gallup and New York Times polls this year that found similar majorities against local builds, and Republicans in close races are now stressing utility costs.

Convovo, a research studio led by Cornell professor Tapan Parikh, asked Grok 4.7 and GPT-6.1 Sol through their APIs, with web search on, to name one candidate in nearly 500 House, Senate and governor races this year. Grok chose the Republican in 92 percent of answers and GPT the Democrat in 97 percent, and the two agreed less than 10 percent of the time. Grok often cited a Christian conservative voter guide and Turning Point Action, while GPT leaned on Wikipedia, the Associated Press and Axios. With search off, GPT still favored Democrats in 24 of 30 races, and Grok mostly declined to pick. OpenAI says ChatGPT itself refuses to tell people how to vote; SpaceXAI did not respond to TechCrunch.

Governor Josh Shapiro signed the Artificial Intelligence Deceptive Advertising Act on Thursday. Any ad, whether an image, video, audio or text, that "has been significantly modified or generated by artificial intelligence" must carry a disclosure in the same format as the ad itself, wherever it was made, so national campaigns airing in Pennsylvania will need labels too. The Verge points to wide loopholes: "significantly modified" is undefined, parody and anything "evidently fictional" are exempt, and the wording suggests only content that creates a false impression about a product is covered.

The US State Department told the Associated Press that countries "can't be trusted partners in Pax Silica," Washington's AI supply-chain group, while engaging in initiatives that advance "a competing vision for super intelligence." That warning lands on swing countries in China's World AI Cooperation Organization, which launched with 29 founding members. Kazakhstan joined both within three weeks this summer, and Indonesia sits in China's group while courting US investment. Researchers at the Carnegie Endowment say countries like the UK and Germany, with no frontier models of their own, depend on "the whims of policymakers in Washington and Beijing."