In partnership with
Three of the biggest names in AI each closed or chased a multi-billion dollar acquisition this week, and none of the buyers are who you'd expect. A payments company, a rocket company, and an AI lab all decided the fastest way to get ahead is to just buy the competition. Underneath all that dealmaking, two smaller stories show what happens when AI tools get more access than they've earned. Let's start with the deal that could quietly change how every AI tool you use gets built and paid for.
Thinking about hiring globally? Start with an EOR.
The best person for your next role might not live near your office—or even in the same country.
More companies are realizing they don't need to open entities everywhere just to access global talent. Instead, they're using EOR to hire internationally faster, stay compliant, and avoid building local infrastructure before they're ready.
Oyster's EOR helps companies hire, pay, and support employees in 180+ countries while Oyster handles payroll, compliance, taxes, and local employment requirements.
Stripe Just Became the Toll Booth for AI
What happened: Stripe agreed to buy OpenRouter, a startup that routes AI requests across more than 400 different models, for more than $7 billion, according to TechCrunch. That's roughly five times what OpenRouter was worth just three months ago, when its weekly token volume was a fraction of what it is now. The deal turns a company best known for processing your credit card payments into one of the biggest middlemen in AI infrastructure.
Why it matters to you: If you use any AI tool built on top of OpenRouter's routing (a growing number of newer AI apps are), Stripe now has a hand in how it works, what it costs, and who gets access first. Consolidation like this usually shows up as higher prices down the line, not lower ones.
What to do about it: No action needed today. Just don't be surprised if AI tool subscriptions get pricier once Stripe folds this in.
SpaceX Just Closed a $60 Billion Deal for an AI Coding Tool
What happened: SpaceX officially closed its $60 billion acquisition of Anysphere, the company behind the AI coding tool Cursor, according to Yahoo Finance. The price is about 24 times what Anysphere was worth less than two years ago, and it's being called the largest startup acquisition ever. Cursor's tools are now folded into a new SpaceX AI division.
Why it matters to you: Even if you don't write code, this is a signal of how much cash the biggest players will burn to control AI development tools. When a rocket company pays 24 times an old valuation for a coding startup, it tells you AI coding tools are becoming core infrastructure, not a side bet.
Anthropic's CEO Says the AI Backlash Is a "Crisis of Trust"
What happened: After investor Gavin Baker accused Anthropic CEO Dario Amodei of stoking anti-AI sentiment to box out competitors, Amodei fired back in a lengthy post on X, reported by TechCrunch. He called the public's growing distrust of AI "fundamentally a crisis of trust," rejected the idea that the only options are heavy regulation or none at all, and argued that curing diseases, not marketing campaigns, is what will actually win people back.
Why it matters to you: This fight over how much to trust AI companies isn't just Silicon Valley drama. It's shaping the regulations, disclosures, and guardrails that will eventually apply to the AI tools you use every day.
Anthropic Is About to Make Its Biggest Purchase Ever
What happened: Anthropic is reportedly in talks to buy Decart, an Israeli AI infrastructure startup, for about $6 billion, according to Fortune. It would be Anthropic's largest acquisition to date, reportedly beating a higher bid from Nvidia, with the deal expected to be paid mostly in Anthropic stock ahead of its planned IPO this fall.
Why it matters to you: Anthropic, the company behind Claude, is stacking up cash and acquisitions ahead of going public. That usually means faster product changes and pricing shifts as it builds its case to investors.
ChatGPT's New Mac Feature Logs Your Clicks, and Stores Them in Plain Text
What happened: OpenAI rolled out an opt-in feature called Computer History that lets ChatGPT and Codex on Mac remember your past clicks, keystrokes, and app activity so it doesn't need screenshots to follow along, according to The Decoder. Multiple outlets, including TheNextWeb, confirmed that this activity log is stored unencrypted, in plain text, right on your machine.
Why it matters to you: If you or your team ever turn this on, anyone with access to that Mac, or malware that gets onto it, can read a plain-text record of everything you clicked and typed while ChatGPT was watching.
What to do about it: Leave this feature off unless you specifically need it. If you do turn it on, don't do it on a shared or work-issued computer.
A Chinese AI Lab Released a Coding Model That's Unusually Good at Hacking
What happened: Z.ai released GLM-5.3, a new open coding model it says is the strongest of its kind, and the model also topped a leading cybersecurity benchmark, according to VentureBeat. It reportedly already found a real, previously unknown vulnerability in Cursor, the same coding tool SpaceX just bought.
Why it matters to you: AI models are getting good enough at finding software vulnerabilities that the same skill cuts both ways, defense or attack. It's one more reason to keep your business software patched and updated instead of assuming nobody's looking for the holes.
Apple Built a Separate AI Model Just for China, With Help From Alibaba
What happened: Apple became the first foreign company approved by Beijing to offer its own AI model inside China, building it with support from Alibaba, according to MacRumors. Alibaba's Qwen model already powers some Apple Intelligence and Siri features for Chinese users, while Baidu handles search.
Why it matters to you: Even the biggest tech companies run different AI rules and different AI partners depending on the country. If you ever expand into international markets, assume the AI tools you rely on here won't work the same way, or at all, somewhere else.
Someone Hid Secret Instructions for AI Inside a Court Filing
What happened: A plaintiff tried to sneak invisible text into a legal filing that instructed any AI reviewing the document to rule in their favor, a tactic caught because of strange spacing in the text, according to 404 Media. The move backfired and drew scrutiny instead of a favorable ruling.
Why it matters to you: If your business uses AI to review contracts, filings, or any document someone else sent you, assume that document could contain hidden instructions aimed at the AI, not at you. Read whatever the AI flags with a skeptical eye.
The Bottom Line
Add it up and you get a pretty clear picture. The biggest AI companies are buying each other faster than regulators can blink, while the tools built on top of all that money are quietly logging your keystrokes and getting fooled by hidden text in a PDF. None of this means you should panic or unplug. It just means the "trust the AI by default" era is officially over. Read the fine print, keep your team's tools patched, and don't hand any AI tool more access than the job actually requires.
Enjoying the Ride?
If this issue saved you from one bad AI surprise, forward it to one person who needs to see it. New here? Subscribe free and get the next issue tomorrow.
Talk tomorrow,
Mark Shilensky
Follow me personally on Social Media:
Facebook: https://www.facebook.com/MarkShilenskyPage
Instagram: https://www.instagram.com/markashilensky/
Threads: https://www.threads.com/@markashilensky
LinkedIn: https://www.linkedin.com/in/shilensky/
X: https://x.com/markshilensky
TikTok: https://www.tiktok.com/@mark.shilensky
BlueSky: https://bsky.app/profile/markshilensky.bsky.social