Key points
- Nadella endorses "deliberate pacing" and "embedded evaluators"
- Microsoft's MAI code of conduct comes Monday
- Amodei proposed, Altman committed, Nadella consults
- He never names Anthropic or Amodei
Microsoft (MSFT) CEO Satya Nadella endorsed deliberate pacing in AI development on Sunday, using the same terms Anthropic CEO Dario Amodei used a day earlier. He also said Microsoft will publish a code of conduct for its own AI models on Monday.
It came in a post on X at 3:36 p.m. Eastern. Within three hours, it had about 850,000 views. It wasn't a reply to anyone.
Nadella used the terms "deliberate pacing" and "embedded evaluators" without naming Amodei or Anthropic. Amodei's Saturday essay is titled "We Must Pace the Frontier," and its first proposal is a team of outside evaluators with employee-like access at every frontier AI lab, which the essay calls embedded evaluators.
What Nadella actually committed to
One thing, with a date. Microsoft will publish "the 'Code of Conduct' that underlies our own first party MAI models" for public consultation on Monday, September 14. MAI is the name Microsoft uses for the models it builds itself. He didn't say what's in the document, how long the consultation runs, or who's invited to respond.
The rest of the post is principle, not commitment. He opened with a condition. "Any pursuit of superintelligence has to be grounded in the core principle that if the AI we build is not helping humanity and under human control, it's not worth pursuing," he wrote.
The endorsement is in paragraph four. "We welcome the research, focus, and deliberate pacing needed to get alignment right as the design goal," he wrote. "We also welcome ideas like 'embedded evaluators' and the broader efforts to develop the mechanisms to make this more than just talk."
Then a limit on who runs it. The effort "cannot be controlled by a handful of entities, but must have broad representation across the ecosystem, countries, and fields, including academia."
He summed up Microsoft's approach as "broad access and choice at every layer of the AI stack; enterprise control of learning loops and models." Whether Monday's document says anything about evaluator access, capability thresholds, or pacing Microsoft's own training runs, the post doesn't say.
How the three positions differ
Three statements in two days, and they're three different kinds of thing.
Amodei made a detailed proposal. "We must slow the pace at which we improve the capabilities of AI models," he wrote Saturday morning. Anthropic committed on its own to give outside evaluators permanent, employee-level access, and Amodei asked governments to require the other frontier labs to match it. His essay lays out three tiers: companies acting alone, then industry coordination inside democracies with government backing, then a global effort led by the US and other democratic governments. We covered the details on Saturday.
OpenAI CEO Sam Altman made a company commitment that afternoon. "I agree with Dario that we need to pace the frontier," he wrote. "Committing to having independent evaluators with employee-like access is a great idea, and we will do the same." No evaluator named, no date, and "we'll have more to share soon."
About 27 hours after Altman, Nadella promised a consultation document. He endorsed the ideas and set no evaluator terms, no pacing commitment for Microsoft's own models, and no governance structure beyond saying it can't be controlled by a handful of entities.
There's a second difference, and it's about Microsoft's own business. Paragraph three of the post is addressed to companies that use AI rather than build it. "Every organization should be able to build its own continuous learning loop/hill climbing machine, without becoming dependent on any one model provider," he wrote, "and have the ability to embed its own knowledge into models and weights they control."
Microsoft is OpenAI's largest outside investor. Under the definitive agreement announced October 28, 2025, Microsoft's stake was roughly 27 percent of OpenAI Group on an as-converted diluted basis after the recapitalization. That agreement extended Microsoft's IP rights to OpenAI models and products through 2032, kept Azure API exclusivity until AGI, and kept the revenue-share agreement in place until an expert panel verifies AGI.
Then on April 27, 2026, the two companies amended it. Microsoft's license still runs through 2032 but "will now be non-exclusive." Microsoft "will no longer pay a revenue share to OpenAI," while OpenAI's payments to Microsoft continue through 2030 under a cap. OpenAI "can now serve all its products to customers across any cloud provider," with Microsoft staying its primary cloud partner and OpenAI products shipping first on Azure unless Microsoft chooses not to support them.
So Microsoft has been loosening its own dependence on one model provider since April, and building MAI alongside it. The Monday document covers those models, not OpenAI's.
