One letter, and the most advanced AI went darkOn 12 June, the US government sent Anthropic a letter. It ordered the company to block its two most advanced models from every foreign national on earth, citing national security. Anthropic could not screen users by passport, so it switched both models off for everyone. No warning, and no explanation of the actual concern. For a couple of years I have talked about this with clients as a risk. A thing that could happen one day. This week it stopped being hypothetical. And if it can happen to Anthropic, it can happen to any US provider. The tools running our hospitals, our banks, our offices, all of it sits on platforms a foreign government can switch off with a single signature.
1 / What actually happenedThe stated reason was national security, tied to a reported way of getting around the model's safety controls. Anthropic disputes that, saying the weakness was minor and common to other tools. You can argue the rights and wrongs of that for weeks. What matters for the rest of us is simpler. A company with a strong safety record, working closely with the same government, was given hours to comply and had no choice but to go dark. 2 / This is not really about one companyIt would be a mistake to read this as an Anthropic story. The lesson is about dependence. Most European businesses now run something important on a US platform, often without a backup. When one provider in one country holds that much of your operation, you have a single point of failure you do not control. This shutdown just made that risk visible for a few days. 3 / Pipe dream or overdue priorityThe honest case against European AI is real. Building it is slow and expensive, the US labs are years ahead, and plenty of smart people say the gap is already too wide to close. The case for it is just as real. AI built in Europe, under European law, is far harder for another government to unplug. Mistral's chief recently warned that Europe has roughly two years to build its own AI base or accept being dependent on someone else's. My own view sits in the middle. Picking a side in a trade fight misses the real problem. Too much of what European businesses run on now depends on a switch they cannot reach. 4 / What a small business can actually doYou are not going to build a national champion. You can do three smaller things. Keep a European tool tested and ready, so you are not learning a new one in a panic. Run your most sensitive work on a model on your own machine, where no one can take it away. And write down, in one page, how your business would keep going if your main AI tool vanished tomorrow. That page is worth more than any prediction about who wins the AI race.
The continent-sized version of this question is hard, and I will not pretend to settle it in a newsletter. Your version is smaller and more useful. Keep a second tool tested and ready, and run your sensitive work where no one can pull it. Then the next government letter is something you read about, while your business keeps running. Talk soon, Mikkel P.S. Hit reply and tell me one US tool your business could not run without for a week. That is the door worth checking first.
The AI Solopreneur • AI Strategy for European SMBs |
