SpaceX Is Officially Buying Cursor for $60 Billion
Days after its IPO sent its valuation to $2.6 trillion, SpaceX is turning around and spending $60 billion on Cursor, the AI coding assistant. The stated goal is winning enterprise customers and closing the gap with Anthropic and OpenAI — which is an interesting strategy for a company whose core product is, you know, rockets. At this point Elon Musk’s portfolio of companies is less a business empire and more a Pinterest board for a man who cannot stop clicking “buy.”
Inside the Fight Over Claude Mythos 5
While America was celebrating a World Cup win, Anthropic was fighting off a Trump administration export control directive demanding suspension of Mythos 5 and Fable 5. This is the follow-on story to the week’s ongoing saga about China, access concerns, and government pressure — and now we have the full picture of just how tense that Friday standoff really was. Whether you think export controls on AI models are necessary national security policy or bureaucratic overreach probably tells you a lot about where you land politically, but either way, this is the new normal for frontier AI companies.
Anthropic’s Latest Feud With the Trump Admin May Actually Help It, Sales Data Suggests
Ramp’s spending data suggests Anthropic’s enterprise sales are growing robustly — and the very public clash with the administration may be functioning as accidental marketing with the tech-skeptical-of-government crowd. There’s a very 2026 irony in the government trying to rein in an AI company and instead turning it into a countercultural hero. If Anthropic’s PR team didn’t plan this, they should claim they did.
‘Dangerous’ AI Models Are Coming No Matter What
Wired makes the uncomfortable argument that the export control fight over Mythos 5 and Fable 5 obscures a deeper truth: advanced AI with serious offensive cyber capabilities is going to be widely available regardless of what any one government does. It’s a clear-eyed piece that doesn’t celebrate the fact — it just refuses to let the policy theater substitute for a harder conversation about where this is actually heading.
Critical Copilot Vulnerability Allowed Hackers to Steal 2FA Codes From Users
A newly disclosed exploit dubbed “SearchLeak” let attackers use Microsoft Copilot to intercept two-factor authentication codes — which is the kind of sentence that should make every CISO sit down slowly. Ars frames this correctly as a systemic failure: the security industry keeps patching individual LLM exploits without reckoning with the architectural reasons these attacks keep working. Adding a powerful AI assistant to enterprise software is exciting right up until it becomes the biggest attack surface in the building.
It Is Trivially Easy to Use Reddit to Manipulate AI Search, Research Suggests
Researchers found that just 13 words of planted text on Reddit, Wikipedia, or Quora can reliably redirect AI search agents toward spam or scam outputs. Thirteen words. That’s less text than this sentence. Given how aggressively AI search tools are being positioned as the future of how people find information, the fact that the manipulation threshold is this low should be a five-alarm fire — not a footnote in an academic paper.
‘Pretty Crazy’ Token Usage Is Testing Bosses’ Bet on AI
Wired got rare candid numbers from real companies about what agentic AI actually costs in practice — and the word “tokenomics” is now apparently something CFOs have to care about. One Silicon Valley software firm described token consumption as “pretty crazy,” which is a polite way of saying the ROI math is getting complicated fast. The gap between “AI will save us money” in the pitch deck and “where did our compute budget go” in the quarterly review is becoming a very real management problem.
DOJ Lawyers Argue xAI Is ‘Vital’ for National Security in NAACP Lawsuit
The Justice Department has intervened in a lawsuit over xAI’s polluting gas turbines — arguing the company is so integral to military operations, including the Iran War, that the case should be dismissed. So to recap: a company that runs AI chatbots is now a national security asset immune from environmental accountability. The “national security” label is doing a lot of heavy lifting in Washington these days, and apparently it now covers Grok.
Bottom Line
The AI industry is discovering simultaneously that its models are national security assets, its costs are blowing up CFO budgets, its search products can be hijacked with a sentence, and its biggest controversies are accidentally the best marketing — it’s a lot to process before lunch.