Nadella backs AI pacing, and picks a fight over who referees
The pacing camp got its fourth member tonight, and the newest one arrived with an objection attached: Microsoft's CEO agrees the frontier should slow down, and does not agree that Anthropic, OpenAI and Google should be the ones deciding what that means.
Satya Nadella wrote on X shortly after 7:30 PM ET that Microsoft "welcome[s] the research, focus, and deliberate pacing needed to get alignment right as the design goal," and announced a "Code of Conduct" governing Microsoft's own first-party MAI models, to be published tomorrow for public consultation. The post is the first time Microsoft's chief executive has put himself inside the framework Dario Amodei proposed Saturday and Sam Altman matched Sunday — and unlike both, Nadella names the mechanism he will accept. He welcomes "embedded evaluators," the third-party reviewers Amodei wants sitting inside labs with badges and company laptops, but ties them to a condition: governance "cannot be controlled by a handful of entities, but must have broad representation across the ecosystem, countries, and fields, including academia."
The sentence is aimed at three labs in a room, and it is aimed at the open-source question. The Information reported today that Anthropic, OpenAI and Google have held working-group meetings since July on an industry-led standards body. Microsoft has no seat at that table and is using the pacing debate to argue there should not be one it is excluded from. The same logic runs through the rest of the post: Nadella says a frontier ecosystem needs "both closed and open-source models" to thrive, and that every organization should be able to build its own "continuous learning loop" — his phrase — and "embed its own knowledge into models and weights they control," without depending on any single model provider. That is Microsoft's enterprise franchise restated as a governance principle: the customer, not the lab, owns the weights.
His lab's own claims are the part worth pressure-testing. Microsoft AI describes its MAI family as trained from scratch, with no distillation from other labs and no unlicensed data — a line that lands differently in a week when Anthropic publicly accused Alibaba, Moonshot and DeepSeek of running Claude through their training pipelines. Microsoft also says a model tuned for Excel matches GPT-5.4 while being up to ten times more efficient, and that a Mayo Clinic co-created clinical model will be owned by Mayo Clinic. All of that is the company's own reporting, and the safety and technical reports it says it published alongside are the place to check it.
The read: this is a real widening of the coalition, and a real dispute inside it. Three labs asking to be regulated, and Washington saying no, was one story. Four labs, one of them openly arguing about the franchise — that is a negotiation.
What to watch: whether the code of conduct Microsoft publishes Monday binds its own models to anything measurable, or reads as a consultation document. Amodei's essay named no pacing threshold either.
We tracked the industry half of this shift two days ago — Altman matches Amodei's evaluator pledge, Musk says 'Dario is right' — and Washington's refusal this morning — Mike Johnson: AI safety is the companies' job, not Congress's.
Microsoft wants a standards body with academia and many countries in the room, and three labs already meeting without it. Is broad representation the fix, or the delay? Tell us in the comments.
