Microsoft CEO Satya Nadella: superintelligence must remain under human control

Microsoft CEO Satya Nadella used a Sunday post on X to lay down a very specific marker on superintelligence: if AI cannot help humanity and stay under human control, it should not be pursued at all. He framed the goal as something bigger than raw capability. The point, in his view, is to build systems that serve people rather than outrun them.

He also pushed a broader argument for how AI should be built and used. Nadella said the benefits need to be spread across countries, communities and companies, and that this requires what he called a frontier ecosystem where both closed and open-source models can thrive. In his telling, no single model family gets to dominate the whole conversation.

That theme carried into the enterprise side of the speech. Nadella said companies should keep control of their own knowledge instead of getting stuck depending on one provider, and that they should be able to build systems around their own data. He described that as a way for firms to preserve their unique and tacit knowledge, with their own learning loops and models they control.

He also backed a slower, more deliberate path on alignment, saying Microsoft welcomes the research and pacing needed to get that right. He pointed to ideas like embedded evaluators and said the goal is to make safety mechanisms more than just talk. And he made clear that advanced AI development should not sit with a handful of entities, but should include broad representation across ecosystems, countries, fields and academia. Microsoft, he said, is aiming for broad access and choice across the AI stack, enterprise control of learning loops and models, and a code of conduct for its first-party MAI models that it plans to publish for public consultation.