AI won't end you. Your email provider might

Last week, I wrote about the Anthropic export control shenanigans and vendor lock-in. Then Cory Doctorow’s newsletter showed up and took the idea a step further into the light.

His point: when Greenland became the target of Trump’s hunger for power, the scary scenario wasn’t Denmark losing its ChatGPT subscription. It was Microsoft reaching into a ministry’s accounts and bricking the lot: email, calendars, contacts, … all gone. Because one company, on the other side of the pond flipped a switch on request.

That’s not a hypothetical about AI, but it is one about every phone, ERP system, identity platform, and even tractor with a kill switch and an owner who answers to a different government than you do.

A lot of digital sovereignty panic these days is aimed at AI, because that’s the thing we’ve all been told to be scared of: can we train our own European model? What if we lose our chatbot? Etc.

But, honestly, yank a country’s ChatGPT or Claude access tomorrow and pretty much nothing will fall over. No hospital depends on it, no factory floor will grind to a halt, no railway will stop functioning. Yank an email provider, a phone fleet, or the software running infrastructure and you’ve got a crisis, fast.

So, yet again, stop asking the wrong questions. Not: “what’s our AI strategy” but “which of our dependencies can be switched off by someone outside our jurisdiction, and what happens the day they do that”.

Same story I was circling with the export control email: worry about who’s got their thumb on the kill switch for the infrastructure you actually depend on. You’ll live without AI.

Colin