Good morning,

As you might know - this newsletter hails from Belgium, Europe. And I might trigger my American readers this week. But perhaps it’s time they get our perspective.

For years we told ourselves a story over here. That America was the difficult friend. A little loud, a little reckless, but our friend. The one holding the line, the one whose tech we ran our hospitals and banks and newsrooms on because, well, they were the good guys. We grew up with American TV. Like the TV-show they were our “Friends”.

But we never read the fine print. You don't read the fine print with friends.

Our ‘Friends’ gave us the best AI model in history. For 4 days.

Then on June 12 a man in Washington signed a piece of paper, and ninety minutes later the most powerful AI on earth went dark in European hands. Not because it broke. Because someone three thousand miles away decided we were no longer allowed to hold it.

Not just us. The British. The South Koreans, who make the actual chips the thing runs on, locked out of the thing their own silicon powers. Every ally, every partner, every "special relationship," all filed under the same cold little phrase: foreign nationals.

Let that one land. Foreign nationals. That's what we are now. Not partners. Certainly not friends. Users. And users can be switched off.

The friendship was a transaction and the invoice just arrived.

To be honest, I don't even think they meant it as an insult. That's the part that haunts me. It wasn't malice. It was a national security memo, processed on a Friday evening, with no explanation attached for the people it cut loose. They didn't hate us. They just didn't think about us at all. We weren't in the room. We're never in the room.

We’re not going to get political here but it’s the direction the White House has been taking for the last few years. “America good, rest of the world bad.”

But what nobody in Brussels wants to say into a microphone is that we did this to ourselves. We outsourced our future to a country across an ocean and called it a partnership. We ran a continent on American clouds, American chips, American models, and we mistook this convenience for safety. A friend who can turn off your lights from another time zone was never really a friend. He was a landlord.

The masks are off in the meantime. That much is clear.

The British already get it. The week of the shutdown their AI minister started talking sovereignty and waving a billion pounds at homegrown infrastructure. The French were even earlier. They banned Microsoft from government PCs - everybody to Linux. Brussels has opened an investigation, which is the European way of admitting you've been caught with your trousers down. We are, all of us, suddenly very interested in building our own.

A decade too late. But interested.

Maybe the relationship survives this. Maybe Washington flips the switch back, and we all agree to call it a misunderstanding. Probably we will. We are good at pretending over here.

But I won't forget the feeling. The morning I opened my laptop and learned, in real time, that the most important tool of my working life answers to a government that does not consider me a person worth notifying.

The future is being built right now. The only question that matters is whose finger sits on the off switch.

So what about you? Are you still betting your work, your business, your country, on a wonky friendship?

Or have you finally started reading the fine print?

Let’s get to work.

Welcome to the Blacklynx Brief

Anthropic pulls Fable and Mythos worldwide after U.S. export order Anthropic took its two most powerful models offline globally after the Trump administration ordered it to block all foreign access, citing a disputed jailbreak the company called minor. Days later, 100+ cybersecurity leaders signed a "Free Fable" open letter arguing the ban handcuffs defenders while attackers pull the same capability from rival models. (Anthropic)

SpaceX buys Cursor in $60B all-stock deal SpaceX exercised its option to acquire AI coding startup Cursor entirely in stock, days after a post-IPO rally nearly doubled its value and pushed Elon Musk's net worth past $1T. Cursor's next model will reportedly be trained from scratch and "as big as Opus," handing Musk a frontier-scale developer platform. (SpaceX)

Z AI's GLM-5.2 pushes open weights to the frontier Chinese lab Z AI released GLM-5.2, an MIT-licensed open-weights model that beats GPT-5.5 on coding, reasoning and math benchmarks and lands just below Claude Opus 4.8. It ships with a 1M-token context window at a fraction of frontier pricing — the closest an open model has come to the top closed systems. (Z AI)

Amodei warns regulation is moving at "tree speed" Anthropic CEO Dario Amodei published "Policy on the AI Exponential," urging regulators to gain the power to ground frontier models and screen them across four risk areas as hacking capability turns models into "tools of strategic consequence." His proposals also include planning for mass unemployment, UBI, and tighter chip export controls. (Dario Amodei)

Bezos bets $41B on an "artificial general engineer" Jeff Bezos raised $12B at a $41B valuation for Prometheus, a startup building AI to design and build complex physical machines like jet engines. Bezos also pushed back on AI job-loss fears, predicting the productivity boom will create "more than 10x" the opportunities. (WSJ)

That’s it for us this week. Please like and subscribe 🙂