The Week's Defining Story: Washington vs. Anthropic
The biggest story of the moment isn't a product launch — it's a government order that's reshaping how America thinks about AI exports. Three days after Anthropic released its two most powerful models, Claude Fable 5 and Mythos 5, the Trump administration issued an export control directive on a Friday afternoon ordering the company to cut off foreign user access. The timing — 5:21 PM on a Friday — left little room for response.
What followed has been a fast-moving policy crisis. Anthropic executives flew to Washington on Monday for White House talks, but those meetings ended without resolution. Cybersecurity experts have weighed in sharply, calling the restrictions "dangerous" — arguing that blocking allies and legitimate researchers from frontier AI tools doesn't protect national security so much as it cedes ground to less scrupulous providers. Prediction markets are apparently pricing in a reversal, with bettors eyeing a quick rollback as expert pressure on the administration mounts.
The episode exposes a tension that's been building for months: the U.S. government wants to lead in AI, but the tools that would help it do so keep running into the very export frameworks designed for a different era of technology.
SpaceX Goes All-In on AI Coding
Elon Musk's SpaceX agreed to acquire Anysphere — the company behind the wildly popular AI coding assistant Cursor — in an all-stock deal valuing the startup at $60 billion. The announcement came just days after SpaceX's IPO, and the SEC filing confirming the deal landed almost immediately after.
The price tag is remarkable for a coding tool, even a beloved one. Cursor has become the go-to assistant for software engineers across the industry, and the acquisition signals that SpaceX sees AI-assisted development not as a productivity perk but as core infrastructure. Whether this is a play for vertical integration — keeping Cursor's edge internal — or a broader bet on developer tooling as a business remains to be seen.
Microsoft Cools on Anthropic; ChatGPT Loses Its Majority
Microsoft's AI chief told Bloomberg this week that Anthropic's models are simply too expensive, signaling a meaningful shift in the company's thinking about its AI partnerships. Microsoft has invested heavily across the AI ecosystem, but the public pushback on pricing suggests the company is actively evaluating cheaper alternatives — a development that can't be welcome news for Anthropic as it simultaneously fights a government standoff.
Meanwhile, OpenAI is facing its own market pressure. For the first time, ChatGPT controls less than 50% of the AI chatbot market, according to TechCrunch. ChatGPT remains the dominant single player, but the milestone matters: the market is fragmenting faster than many expected, and OpenAI's early mover advantage is eroding as rivals chip away at its lead.
Meta Moves, Jensen Bets on America
Meta announced a broad wave of AI features coming to Facebook on Monday, led by a new "AI Mode" for search and expanded photo editing capabilities. The rollout is the clearest signal yet that Meta is treating Facebook not as a legacy product but as a delivery surface for its AI ambitions — bringing AI directly into where its billions of users already spend time.
Nvidia CEO Jensen Huang, meanwhile, is making a different kind of bet. Huang argued this week that AI will generate manufacturing jobs in America rather than displace them, and Texas is set to be the proving ground. The claim is contrarian — most forecasts expect AI to reduce factory headcounts — but Huang's reasoning centers on the infrastructure buildout required to run AI at scale, which demands human labor to construct, wire, and maintain.
xAI's Trade Secret Case Against OpenAI Dies
A federal judge dismissed xAI's trade secret lawsuit against OpenAI with prejudice, ending Elon Musk's legal effort to accuse his former colleagues of stealing proprietary information. Dismissal with prejudice means xAI cannot refile the same claims. The suit had been seen by many observers as more strategic than substantive; its permanent closure removes one front in the long-running Musk-OpenAI feud, though few expect the broader rivalry to cool.
The through-line today is friction: between governments and AI companies, between incumbents and challengers, between the promise of AI productivity and the hard questions of who gets to use it, at what cost, and on whose terms.