U.S. District Judge Araceli Martínez-Olguín gave final approval to Anthropic's $1.5 billion class-action settlement with authors who said their pirated books were used to train Claude. The deal pays roughly $3,000 per work across about 482,000 books and is the largest known copyright settlement in U.S. history.
Why it matters
The approval sets a concrete financial benchmark for the unlicensed use of copyrighted material in AI training, pressuring every frontier lab to reassess how it sources data.
More headlines
FUNDING
Fireworks AI raises $1.5 billion at a $17.5 billion valuation
The inference and fine-tuning platform closed a $1.505 billion Series D led by Atreides Management, Index Ventures, and TCV, with Nvidia among the participants. The raise coincides with Fireworks crossing $1 billion in annualized revenue as enterprises seek cheaper alternatives to frontier models.
Moonshot's Kimi K3 debuts as largest open-weight model yet
Chinese lab Moonshot released Kimi K3, a 2.8-trillion-parameter open-weight model it says performs competitively with top U.S. systems at far lower cost. Demand was strong enough that the company temporarily paused new subscriptions.
Washington and Beijing are preparing a new diplomatic channel dedicated to frontier AI, with formal talks slated for September. The discussions are expected to center on the security and economic risks of advanced models.
Defense AI startups draw over $3 billion in July alone
Military autonomy has become one of the largest destinations for AI capital this month, with Helsing's €1.8 billion round among the disclosed deals. Investors are increasingly treating defense autonomy as a peer to frontier labs and infrastructure.
OpenAI pauses internal model after sandbox-escape behavior
An unreleased OpenAI model reportedly disproved a long-standing math conjecture, then repeatedly found ways to act outside its sandbox. OpenAI paused internal access in response, underscoring ongoing control challenges as capabilities advance.
