Three stories worth your coffee this Tuesday: Anthropic put a scary sentence in a government filing, the New York Times went after a browser instead of a chatbot, and Marvell spent a mountain of stock on fiber optics. Let's go.
01 of 03
Anthropic Says Self-Building AI Might Be Right Around the Corner
Anthropic filed an amended S-1 with the SEC on Monday, the paperwork for its IPO, and tucked into the Risk Factors section is a warning about recursive self-improvement. That's the point where AI gets so good at building AI that progress takes off on its own.
The key line: Anthropic's internal estimates for future models "indicate that we may imminently meet our Responsible Scaling Policy's definition of 'dramatically accelerated progress attributable to AI.'" It says no current model has crossed that line yet.
It goes further. The filing admits Anthropic "in the past" did not spot a big jump in its own rate of development "until after it has occurred." It also says it does not expect to safely let Claude design and build the next generations of frontier models "without meaningful human supervision or understanding," and that it is considering changes to its Responsible Scaling Policy to keep humans in control.
This is not out of nowhere. Back on June 4, Anthropic published a blog post called "When AI Builds Itself" saying Claude already writes more than 80 percent of the code merged into its systems, and floating the idea of a global option to slow or pause frontier development.
Here is where I land. An S-1 risk factor is half confession and half sales pitch, and "our product is so powerful it scares us" is a great pitch to investors. But you also don't put stuff in an SEC filing that you can't defend, because lawyers sign off on every word. The line that sticks with me is the one about missing a jump until after it happened. That is the whole ballgame with safety. If the folks building it can't see the curve bend in real time, the rest of us sure won't.
Source: Anthropic S-1/A on SEC EDGAR, via TechCrunch
02 of 03
The New York Times Sues Brave for Being the Middleman
On Monday the New York Times filed a 72-page complaint against Brave Software in federal court in Manhattan. In a separate 69-page suit, the New York Daily News, the Chicago Tribune, and the Orlando Sentinel did the same.
The claim is that Brave built an index of more than 30 billion webpages with a crawler publishers could not identify or block, then sold access through its Brave Search API to AI companies. The suits name five customers: Anthropic, Perplexity, Mistral, xAI, and Cohere. The complaints also go after a Brave browser feature that gets around paywalls, and Brave's habit of blocking publishers' ads while selling its own.
The Tribune group gave examples from Brave's AI answers, including more than 1,200 words lifted from a Tribune story about data center tax breaks and nearly 600 words from a Daily News piece on Central Park horse carriages. The Times says it sent cease-and-desist letters to Brave CEO Brendan Eich in August 2025 and April 2026 and never heard back. It wants an injunction and damages. The best line in the complaint calls Brave "a burglar that misappropriates the content of publishers, and an AI 'fence' that sells stolen content to third parties."
Brave did not comment, but it has argued fair use before, and it's already tangled up with News Corp over the same kind of thing.
What I find smart here is the target. Suing OpenAI or Anthropic directly is a long, expensive fight against companies with bottomless legal budgets. Going after the data supplier in the middle hits the supply chain instead. If a court says reselling scraped news through an API isn't fair use, every AI company buying search data has to look hard at where it came from. And the publishers have a real point.
Source: TheWrap and TechCrunch
03 of 03
Marvell Bets Around $115 Billion on Light, and Wall Street Hates It
Marvell agreed Monday to buy Lumentum in an all-stock deal valued around $115 billion. Lumentum shareholders get 4.5 Marvell shares for each Lumentum share, so the final price floats with Marvell's stock. It's not expected to close until 2028.
Lumentum makes optical connectivity tech, the stuff that uses light to link up the chips that train and run models like ChatGPT, Gemini, and Claude. Marvell CEO Matt Murphy told CNBC it's "a double-down on supporting Nvidia," which, for what it's worth, put $2 billion into each company earlier this year.
Investors did not clap. Marvell fell 13% and Lumentum dropped about 19% on the news. Lumentum's stock had run up more than 450% over the past year, so plenty of people think Marvell paid top dollar right at the peak.
My read is that the logic is fine and the timing is rough. Moving data between chips is a real bottleneck, and Jensen Huang has been beating that drum for a while. But this is an all-stock deal that won't close for two years, during a stretch where chip stocks are already getting hammered. If the AI spending boom cools even a little before 2028, Marvell shareholders are holding the bag on a very expensive bet. I'd want to see some regulators weigh in before I got excited either way.
Source: Yahoo Finance / Moneywise