Anthropic & Claude

Jacob Coxon spent three years on model training at OpenAI and Anthropic. His farewell post on X accuses both of gambling with our lives. The striking part is who agreed with him in public.

People using Kimi sometimes got answers from Claude Opus without knowing it. Anthropic's new threat report lays out how two Chinese labs rerouted customer requests and kept the transcripts for their own training.

GSA has signed a new 27-month OneGov agreement with OpenAI. Instead of a dollar per agency, there's a 50% discount on tokens from October 1. The Anthropic and Google deals expire on September 30.

Roughly 500,000 authors are getting $3,000 per pirated work. Now that the money is moving, publishers and agents are turning up with claims — including on books whose rights reverted years ago.

Per The Information, Anthropic plans to publish its IPO prospectus right after Labor Day, hold an investor day in mid-September, and list in late September or early October. The target: a $1.5 to $2 trillion valuation and more than $60 billion raised - the largest IPO on record.

On September 3, Claude, ChatGPT and Grok all failed within the same hour. At Anthropic, Fable 5.1, Mythos 5.1 and Opus 5 were affected for almost three hours, and Claude was the last to come back. Hacker News spent the afternoon asking why all three went down together.

The Trump administration filed a brief in a Manhattan court backing OpenAI in its fight with the New York Times. Training large language models on copyrighted material is generally fair use, it argues. It's the first time Washington has weighed in on this wave of lawsuits - and Anthropic is in the same boat.

A day after the commerce blueprint, Anthropic follows up with a 25-minute guide. The main lessons: one agent with skills beats a subagent architecture, every UI element is a tool, and no model ever gets to move money on its own.

US Commerce Secretary Howard Lutnick says Anthropic is 'back on the right side.' After export controls, a Pentagon blacklist and a court fight, that is quite a turnaround. And co-founder Tom Brown is suddenly the one in the spotlight.

Just in time for the holiday season, Anthropic ships a ready-made kit for building commerce agents on Claude. One partner reports carts up to 35 percent larger and shoppers 60 percent more likely to buy.

Anthropic is locking in compute at scale – a $35 billion cloud deal with Nvidia-backed Lambda, tied to a data center campus in Texas. Just ahead of a planned IPO, it says more about strategy than about hardware.

Simon Willison diffed the Fable 5.1 system prompt against the old one. The most striking addition: a hard ban on reproducing song lyrics. Days after Sony and Warner Chappell sued.

Same price, same tokenizer, but not a drop-in replacement. If you're moving from Fable 5, you'll trip over forced tool calls, bound thinking blocks, and a model that suddenly rewrites whole files.

35 music publishers accuse Anthropic of illegally grabbing copyrighted songs to train Claude. They want up to $150,000 per work — and a line drawn on principle.

Anthropic is renting roughly $45 billion worth of compute from Nscale — 460 megawatts from a West Virginia data center running Nvidia's Vera Rubin chips. The deal is also a message to Wall Street.

Anthropic is opening Claude for Teachers to U.S. schools and entire districts – as a free Enterprise offering with central management. Plus new skills for the back-to-school rush.

Just ahead of its planned IPO, the backlash against AI is catching up with Anthropic. Per CNBC, more than 75 percent of young respondents don't trust Dario Amodei – and the discontent even shows up in the prospectus.

Anthropic is said to be filing its public S-1 as soon as the end of August. The New York Times reports an IPO could value the company at up to $2 trillion — the largest in history.

With the Model Hardware Standard, Claude gets a way to operate robotic arms, microscopes and liquid handlers directly. The research preview is running in the first labs now – and a Genentech example shows how far it already reaches.

Anthropic is handing 10,000 seats to researchers worldwide – free or heavily discounted. On top of that, the AI for Science program is opening beyond biology. A clear push to get Claude deeper into research.