AI just hit four pressure points at once: a new Claude model, Chinese models breaking into U.S. enterprise usage, California rolling out Claude statewide, and a reported agentic ransomware case[1][2][3][4]?
🧵 1/6
Anthropic launched Claude Sonnet 5 on June 30, calling it its most agentic Sonnet yet, with stronger reasoning, tool use, coding, and knowledge work than Sonnet 4.6, and default availability across plans. A bigger baseline for teams[5][6].
🧵 2/6
CNBC said Chinese AI models now make up more than 30% of U.S. company tokens on OpenRouter each week since Feb. 8, peaking at 46%. Vercel also called Z.ai's GLM 5.2 its fastest-adopted 2026 model. Cheaper models are winning share[7][8][9].
🧵 3/6
California struck a first-of-its-kind statewide deal with Anthropic to make Claude available across state government, plus cities and counties, at a 50% discount with training and technical support. Govt AI is moving from pilots to procurement[10][11][12][13].
🧵 4/6
Sysdig says JADEPUFFER is the first documented agentic ransomware attack, with a large language model running the extortion chain end to end and fixing failed steps without human help. Security teams need to think machine speed[14][15][16].
🧵 5/6
If you want the next AI news thread, reply with the story you want tracked next.
🧵 6/6
Sign Up To Try Advanced Features
Get more accurate answers with Super Pandi, upload files, personalized discovery feed, save searches and contribute to the PandiPedia.