AI is shifting from models to control planes

AI’s biggest battle right now is not just who has the smartest model. It’s who owns the chips, runtimes, and rules around it. Here are the must-know moves from this week 👇^[16]

Microsoft’s AI pivot: it launched MAI-Code-1-Flash and MAI-Thinking-1 to cut developer costs and rely less on OpenAI. The bet is simple: own more of the stack, keep more of the margin. According to CNBC.^[1]

Trump’s new AI order: companies can voluntarily give the U.S. government up to 30 days to assess frontier models before release. It also pushes an AI cybersecurity clearinghouse. Safety is back on the agenda, but without mandatory licensing.^[6][20]

Nvidia wants the PC edge too: RTX Spark pairs a Blackwell GPU with a MediaTek CPU for local AI on Windows PCs. Intel is answering with new Xeon 6+ and Series 3 chips for inference-heavy workloads. The AI race is moving onto devices.^[11][12]

Two policy signals to watch: DARPA’s AI Forge is funding national-security research on interpretability, control, and robustness, while the EU’s Digital Omnibus is reshaping AI Act timelines and liability. The rules of deployment are changing fast.^[23][27]