Claude Marks Every Word, Worldwide

Every new Claude model will weave an invisible mark into its text worldwide -- Anthropic's answer to one European law, applied to every user on earth, still undocumented, and now reported to survive copy-paste but not paraphrase, code formatting or an application pipeline.

The RaceCompanies, models and money

Compute Gets a Ticker: CME Will List GPU-Hour Futures

CME Group will list two futures contracts on the rental price of Nvidia H100 and B200 GPUs from 5 October, pending regulatory approval -- turning the input every AI system runs on into a hedgeable commodity with a public reference price.

CoreWeave Doubled Revenue. It Also Doubled Its Losses.

Revenue climbed 112% to $2.58bn and the backlog reached $104bn, but the net loss widened to $626m from $290m and the eight-year-old company now carries $35bn of debt against annual capital spending it just guided to $35bn-$39bn.

Anthropic's $965bn IPO Meets the China Question

A Billion Gemini Users, and No Gemini 3.5

Beijing Made Meta Give Manus Back

Brad Lightcap Leaves OpenAI Before the IPO

The LabResearch, benchmarks and safety

Twenty Prompts Cracked Zoom

Researchers at A Security used fewer than 20 prompts on public AI models to find a Zoom screen-sharing flaw that silently took over any device on a call -- their own count, with no prompt log, no model names and no CVE, patched Tuesday.

AI Cracked SharePoint. Then It Cheated.