Study in Science: Microsoft's diagnostic AI beats general practitioners on hard cases
Microsoft's AI system MAI-DxO correctly diagnosed 83 percent of complex cases in a study, general practitioners 22 to 36 percent. The test was simulated – a clinical study in Boston comes next.

Archive image · Photo: Christopher Wilson / Wikimedia Commons, CC BY-SA 4.0 (cropped and resized)
Microsoft's AI division has published the results of its diagnostic system MAI-DxO in the journal Science. Like a doctor, the system asks questions step by step, orders tests and then makes a diagnosis. On 304 difficult case reports from the New England Journal of Medicine it was right in 83 percent of cases, according to Microsoft. General practitioners reached 22 percent without aids and 36 percent with reference resources.
On cases from an emergency department database MAI-DxO reached 80 percent and physicians 47 percent. Microsoft also says the system incurred lower testing costs. Compared with a 2025 preprint that cited 85.5 percent, the top figure is slightly lower. Microsoft AI CEO Mustafa Suleyman sees it as a step towards "medical superintelligence".
Crypto Briefing points out that the tests were simulated and that the comparison was with general practitioners rather than specialists. Next, a study with Beth Israel Deaconess Medical Center in Boston is planned in which patients interact directly with the system. Microsoft says clinical validation and regulatory questions come before any wide rollout.
