Artificial intelligence has been promising to transform medicine for years. Google DeepMind and Google Research are now showing what that transformation actually looks like in practice, with a wave of tools and collaborations producing results that range from record-breaking genetic diagnoses to flu forecasting that outperformed dozens of rival systems.
A blog post authored by Yossi Matias and Pushmeet Kohli outlined the breadth of Google’s current health AI work, framing it as a continuation of the momentum generated by AlphaFold, the system that cracked protein structure prediction and became one of the most cited scientific tools of the past decade. The new announcements build on that foundation rather than replacing it.
The tools doing the actual work
One of the most consequential releases is MedGemma, an open medical AI model detailed in Nature Medicine. The model is designed to lower the barrier for researchers who want to incorporate AI into molecular discovery and diagnostics without building their own systems from scratch.
Its performance on chest X-ray classification improved accuracy by 15.5 to 18.1 percentage points over previous models, according to the research.
On the mammography side, a study involving 125,000 women found that an AI system identified 25% of cancers that were missed during routine interval screenings. The same system reduced specialist workload by up to 40%.
The genetic diagnosis work is perhaps the most striking. In partnership with Stanford University School of Medicine, Google’s tools enabled a genetic diagnosis to be completed in under eight hours for time-critical cases, a result that set a Guinness World Record.
On October 15, DeepMind launched Google Antigravity, an interactive AI workbench built specifically to accelerate the identification of antimicrobial peptides, or AMPs. Antimicrobial resistance is projected to cause 10 million deaths annually by 2050, according to the research. Antigravity integrates capabilities from ProtNLM and AlphaFold to help researchers prioritize the most promising peptide candidates, and early findings are already moving into wet lab validation.
Drug discovery is another area showing concrete output. Using AlphaFold to screen 30,000 compounds, Google’s research contributed to the development of two FDA-approved drugs for Chagas disease.
Public health forecasting and what comes next
Beyond the clinic and the lab, Google’s flu forecasting model ranked first among all systems evaluated by the CDC for predicting hospital admissions during the 2025 to 2026 flu season, outperforming 38 competing systems.
A collaboration with Imperial College London is also part of the broader initiative, though the specific outputs of that partnership were not detailed in the announcement.
Google has committed over $60 million in funding for 2026, with an emphasis on open-access models aimed at clinical applications.
Disclosure: This article was edited by Diego Almada Lopez. For more information on how we create and review content, see our Editorial Policy.