Sony Music Publishing and Warner Chappell Music filed suit against Anthropic late Friday in the U.S. District Court for the Northern District of California, naming CEO Dario Amodei and co-founder Benjamin Mann personally as defendants. The 48-page complaint accuses the company of a "brazen campaign of illegally torrenting, scraping, and downloading copyrighted works on a massive scale" to build Claude.
The publishers say thousands of copyrighted compositions — from "Eye of the Tiger" to "All I Want for Christmas Is You" — were harvested and repeatedly copied to train the models, and that lyrics can allegedly still be coaxed out of Claude nearly verbatim. They're seeking a jury trial plus up to $150,000 per infringed work and $25,000 for every instance of stripped copyright-management information — a bill that could run into the billions.
It's the fourth major music-publishing suit against Anthropic, following an earlier case from Universal Music, Concord and ABKCO. This one leans harder on piracy: it cites a separate court's description of Anthropic's book-training practices as "straightforward piracy but at massive scale" — the same finding that led to a $1.5 billion settlement in the Bartz authors' case earlier this year.
OpenAI told Cursor it will terminate its model-access contract on November 12, 2026 — the latest date allowed under its agreement — after SpaceX completed its $60 billion acquisition of the coding startup earlier this month. OpenAI invoked a change-of-control clause and said it "cannot be confident that SpaceX will use our technology within our terms of service, based on our experience with Elon Musk's companies violating contracts."
Cursor co-founder Michael Truell downplayed the impact, noting OpenAI models account for roughly 5% of Cursor's user traffic; Claude, Gemini and Grok remain available inside the tool. Musk responded on X that he "couldn't care less," reviving his long-running claim that Altman betrayed OpenAI's nonprofit roots.
Cursor had reached a $29.3 billion valuation and over $1 billion in annualized revenue before SpaceX more than doubled that figure with its buyout — a reminder that in AI coding, owning the wrapper is no longer enough if a rival controls the model underneath.
a16z closed a $1.1 billion Machine Age Fund dedicated entirely to hardware — chips, memory, networking, cooling, data centers, robotics and even home AI appliances. Five senior partners, including Ben Horowitz and Martin Casado, put their names on the launch, an unusually heavy commitment for a firm best known for backing software.
The thesis is physical, not financial: compute density has jumped 28-fold from Nvidia's H100 generation to its Rubin racks, and a rack that once drew 5–10 kilowatts now draws up to 250. Hardware has climbed from a marginal slice of a16z's deal flow to more than 20% of it, with the portfolio already spanning Skydio, SpaceX, Anduril, Waymo, Nexthop and the Rivian spinout Mind Robotics.
It's a bet that the next constraint on AI progress isn't better algorithms, but enough silicon, power and steel to run them — a view increasingly shared across the industry as memory and compute shortages ripple into consumer products.
The Trump administration is reportedly weighing a broader round of semiconductor tariffs that would reach past raw chips into the finished products built with them — laptops, gaming consoles and data-center servers — according to Politico, citing eight people familiar with the discussions. The plans remain preliminary, with a phase-in period under consideration rather than a single cutover date.
Commerce Secretary Howard Lutnick reportedly favors tying tariff relief to a company's investment in U.S. chip manufacturing. The White House says reshoring semiconductor production remains a top priority and points to "hundreds of billions of dollars" in investment already secured.
Tech companies have begun lobbying against the plan, warning that stripping January's data-center carve-outs could raise the cost of the very AI infrastructure the U.S. is racing to build, with one analysis putting the potential GDP hit near $90 billion a year.
ChangXin Memory Technologies confirmed its LPDDR6 memory has entered mass production, and the upcoming Xiaomi 18 Fold will be the first phone in the world to ship with it — ahead of equivalent modules from Samsung and SK Hynix, which have announced but not commercialized the standard. Xiaomi's own Xring O3 chip, unveiled days earlier, is the first mobile silicon built to support it.
The new memory reaches speeds of up to 12,800 Mbps with capacities up to 16GB per chip, improvements CXMT says are specifically aimed at heavier on-device AI workloads and multitasking rather than raw benchmark bragging rights.
For a company that only reached LPDDR5 mass production last year, leapfrogging established suppliers to a flagship launch narrows the gap with global memory leaders — and arrives as a persistent global memory shortage keeps pushing device prices upward.