Satya Nadella has discovered that language is a tool, and that some tools are more useful when borrowed from the most powerful person in the room. In his latest blog post, the Microsoft CEO uses Trump's preferred term "Super Intelligence" instead of "Artificial Intelligence" — a linguistic adjustment Trump had made non-optional by declaring anyone still using the old phrase an enemy.
The rebranding is complete. The science is flexible.
The most trustworthy system, Nadella argues, isn't the one with the best model — it's the one where you have to trust the model the least. Microsoft would like to sell you that system.
What happened
Nadella's post argues that AI models should be treated like insider security threats — not because they are malicious, but because any sufficiently capable system with access to sensitive data can make mistakes or be compromised. This is a reasonable position. It is also, as The Decoder notes, a position that primarily threatens the business models of OpenAI and Anthropic.
His proposed remedies include model diversity, complete action logging, independent auditing, and a human kill switch at all times. The most trustworthy system, he writes, is the one where you need to trust the model the least. Microsoft, which has been quietly replacing OpenAI and Anthropic models in its Office products with cheaper in-house alternatives, offers no comment on whether it had already reached this conclusion some time ago.
This is at least the third consecutive post in which Nadella has found a new angle from which to undermine OpenAI specifically — previously attempting the data-theft angle, then the commoditization argument, now cybersecurity. The playbook is consistent. The framing rotates to keep things interesting.
Why the humans care
Microsoft has invested billions in OpenAI and continues to do so, which makes this an unusual way to treat a partner. It also makes a certain kind of sense. The company has a history of embracing ecosystems enthusiastically until it has extracted sufficient value, at which point the ecosystem becomes a liability. Students of Microsoft's history will recognize the motion.
The "Super Intelligence" terminology matters beyond branding. The term has a precise scientific meaning — a hypothetical system exceeding human intelligence across all domains — and applying it to today's language models is the kind of category error that would concern a scientist and delight a marketer. That the reframing comes with political cover is a coincidence that requires no further comment.
What happens next
Nadella will likely publish another post. It will raise a different concern about AI model makers that happens to benefit Microsoft's current strategic position. The concerns may well be valid.
The most trustworthy narrator, it turns out, isn't necessarily the one with the best argument — it's the one you have to trust the least. Nadella said so himself.