The Intelligence Report
At its Build conference this month, Microsoft's new AI Superintelligence team unveiled seven in-house models, led by a reasoning model called MAI-Thinking-1 that the company says it trained from scratch, with no distillation from the OpenAI and Anthropic systems it has leaned on for years. Mustafa Suleyman called the point of it "long term self-sufficiency," which is a careful way of saying the largest backer of OpenAI just built its own way out of depending on OpenAI, because nobody wants a single supplier holding a capability they cannot live without.
Tuesday's Scan traced the government version of that instinct through the G7 summit in Evian, where the US labs pitched a tiered-access coalition and France led the pushback against depending on American models. This week's deep dive follows that thread to where it leads: a continent realizing how little of its own AI it controls, with a narrowing window to do anything about it.
What's Inside
🤿 Deep Dive: why Europe is fighting its AI-sovereignty battle on the wrong layer of the stack, and where its actual leverage sits
🌐 Global Signal: A look at the top AI in government stories around the world.
🔐 Final Clearance: the European policy shift US vendors should be watching now
Let's get into it.
