Microsoft CEO Satya Nadella has backed external evaluators and deliberate pacing in AI development. His September 13 original post says AI must benefit humanity and remain under human control.
Nadella welcomes embedded evaluators but calls for broad participation across companies, countries and academia. Control should not rest with a handful of entities.
He connects this to choice between open and closed models. Organizations should retain control over their knowledge and learning processes without depending on a single model provider.
The post also announced Microsoft’s own code of conduct for public consultation. It does not pledge a joint development pause or establish that independent evaluators were already appointed. METR now links it in its October proposal as earlier support for embedded evaluation.