Microsoft has formed a new Superintelligence team inside its AI division, tasked with developing what the company calls “humanist superintelligence” — advanced AI systems that remain under human control. Mustafa Suleyman, CEO of Microsoft AI, announced the team Thursday morning in a blog post, and will lead it himself, with chief scientist Karén Simonyan and several researchers Microsoft had already poached from Google, DeepMind, Meta, OpenAI, and Anthropic.

The move comes after Microsoft’s landmark OpenAI deal had barred it from pursuing its own AGI research and capped how large a model it could train. Suleyman tells Fortune the company now has, in his words, a “best-of-both environment,” where it is free to pursue its own superintelligence and work closely with OpenAI.

Suleyman’s vision explicitly rejects the open-ended race to AGI. Instead, he writes, Microsoft will build domain-specific systems that improve human well-being, with early focuses on medical diagnosis and clean energy. He cited Microsoft’s MAI-DxO model, which solved 85% of diagnostic challenges from The New England Journal of Medicine’s case studies, compared to human doctors’ roughly 20%.

Rivals are already positioning themselves around the same idea, with Meta rebranding its AI efforts, OpenAI looking beyond AGI and Anthropic researching control of a hypothetical future superintelligence.