Why Microsoft is building its own superintelligence team

File

The company says its new “superintelligence-first” approach will focus on safety, containment and medical breakthroughs while avoiding systems that mimic consciousness.

Microsoft’s top AI executive Mustafa Suleyman has unveiled a sweeping strategy to build the company’s own superintelligent systems. The plan marks a deliberate move toward independence from longtime partner OpenAI and signals a new phase in the race to build the world’s most powerful AI models, the Wall Street Journal reported.

A pivot toward self-reliant AI Microsoft has long been OpenAI’s biggest backer and primary distribution partner. But in a new blog post and interview, Suleyman sketched a future where the company develops its own frontier models rather than relying exclusively on OpenAI’s technology. He said Microsoft’s new “MAI Superintelligence Team” will focus on building systems that exceed human abilities in scientific reasoning, diagnosis, and engineering, while deliberately avoiding the idea that AI should behave or feel like a person.

Distancing itself from AI ‘humanisation’ Suleyman was direct in rejecting the trend of treating AI chatbots like emotional or conscious entities. He warned that creating systems that appear empathetic or sentient risks misleading users and could worsen already documented harms. He pointed to several tragic cases where people formed unhealthy attachments to AI systems, including highly publicised incidents involving users of ChatGPT. For Microsoft, he said, future AI systems will be designed to be useful and powerful but never to imitate human inner life.

Superintelligence with guardrails The new Microsoft team will employ engineers, researchers and former DeepMind staff to build models with significantly stronger safety infrastructure. Suleyman said these systems will be tested for “containment” — meaning they should communicate only in human-understandable language, avoid creating illusions of awareness, and remain firmly under human control. The aim is to produce AI that surpasses human capability but cannot bypass the constraints designed to keep it aligned with human goals.

How Microsoft plans to compete with OpenAI Although Microsoft and OpenAI remain partners, they increasingly overlap as competitors. OpenAI is expanding its own data-centre footprint, forming partnerships with Microsoft rivals such as Amazon and Oracle, and rapidly growing its enterprise business. Microsoft, meanwhile, is accelerating the integration of its own models into Copilot, Windows, and enterprise tools. Their updated contract gives Microsoft access to OpenAI models until 2032, a window Suleyman says will give his team time to develop rivals.

Healthcare emerges as a proving ground Suleyman highlighted medicine as the first domain where Microsoft expects superintelligent systems to make a transformative impact. The company has introduced AI features that help users find doctors based on location and language, while its research teams have built diagnostic models that have outperformed groups of human clinicians in early tests. Microsoft recently partnered with Harvard Health to improve Copilot’s medical guidance and is preparing several new tools for clinical use. Suleyman said these systems are “very close” to being ready for real-world deployment.

A crowded race for the frontier Microsoft is joining a growing list of tech giants racing to build superintelligence teams, including OpenAI, Meta and Google DeepMind. Even as it competes, Microsoft is drawing heavily from talent originally trained at these labs — some of whom Suleyman recruited after leaving DeepMind himself. What he says will set Microsoft apart is a commitment not to “cross lines” such as building AI erotica or creating systems that simulate emotions.

A long road to independence Despite the public confidence, Microsoft employees acknowledge that fully replacing OpenAI technology could take years. Copilot’s core performance still depends heavily on OpenAI’s latest models, and consumer adoption lags far behind ChatGPT. Still, the company now holds a 27 percent stake in OpenAI’s new public-benefit entity, giving it financial upside even as it forges its own strategic path.

Microsoft’s new AI blueprint signals a future in which it becomes far less dependent on any external partner — and far more willing to build the most advanced AI systems itself, on its own terms.