Google Claims Tangible Gains from Its AI in Health and Climate

Google highlights the deployment of AI tools in crisis management, biology, and education, complemented by usage and accuracy statistics. The company also announces new models and partnerships, emphasizing the need for collective and responsible implementation.

Forecasting and Alerts: PPE, Monsoons, Floods, and Fires

Google describes the Earth AI Planetary Prediction Engine as an autonomous system capable of quickly anticipating crises with natural language instructions. The company claims that this tool identified 83% of hotspots in advance during an Ebola outbreak in the Democratic Republic of the Congo, doubled the accuracy of food security forecasts in Nigeria, and outperformed traditional models in the United States for identifying vulnerable communities based on 21 CDC indicators. Google indicates that it continues to expand the forecasting horizons of this system. By 2025, the company reports that its monsoon forecasts provided information to 38 million farmers in India. It states that Flood Hub now covers 2 billion people in over 150 countries and includes river floods and flash floods. Regarding wildfires, Google cites assistance in predicting perimeters in the United States and 33 other countries, and collaborations with partners towards a FireSat constellation to detect fires too small to be seen before. In 2025, over 520 crisis alerts related to fires were disseminated on Google Search, providing information to more than 75 million users. Google also recalls the use of WeatherNext by Jamaican authorities to anticipate the trajectory of Hurricane Melissa and the activation of a seismic alert system that warned millions in Venezuela, while reporting progress for cyclones and extreme heat. The WeatherNext 3 version combines real-time satellite observations and AI for high-resolution hourly forecasts without a massive supercomputer, presented as a breakthrough for low-data-density regions. Google describes WeatherNext 3 as the most advanced and accurate global model, with precipitation forecasts announced 50% more accurately a day or more in advance, and already integrated into its products.

Health and Biology: Screening, Genomics, and Research Tools

A Google AI detected 25% of interval cancers missed in mammograms involving 175,000 women, according to a study conducted with Imperial College London and the NHS in the UK. The company reports advancements in tools for detecting lung cancer, colorectal cancer, and tumor mutations. It notes that 40% of people with tuberculosis are undiagnosed and mentions over 25,000 X-rays screened via its system used by Nexus Intelligence, across 40 sites in six countries, as well as bioacoustic models leveraging coughs through Health Acoustic Representations. Its diabetic retinopathy model is said to have supported over 1.15 million screenings, with Google aiming to reach 6 million in the next decade. In structural biology, Google presents AlphaFold as a Nobel Prize winner that has predicted 200 million protein structures, with 4 million researcher users in 190 countries, for applications ranging from drug discovery to studying neglected diseases like Chagas disease and leishmaniasis. AlphaMissense is described as helping to predict pathogenic mutations. Google claims to have mapped 9 billion single nucleotide variations through AlphaGenome Atlas and made it accessible, stating it is developing a new capability to predict the effect of variations on cellular behavior. Google highlights Co-Scientist to accelerate the generation and validation of hypotheses, particularly to identify potential therapeutic uses of drugs for acute myeloid leukemia. The company has launched DeepConsensus, DeepVariant, and DeepPolisher, which it credits with contributions to the completion of the human genome, the drafting of the first pangenome, and the work of the Human Pangenome Reference Consortium. Google specifies that prospective validations are being conducted in real-world conditions with partners, including Beth Israel Deaconess Center, via AMIE, and that a nationwide trial is underway, while also highlighting support for frontline caregivers and the time freed up for doctors. Additionally, the company mentions the development of smartphone and connected device solutions to enable early detection of cardiovascular diseases, insulin resistance, hypertension, absence of pulse, as well as passive monitoring of heart rate, and discusses a collaboration in Arkansas aimed at improving health in rural areas.

Languages, Economy, and Aviation: Expanded Coverage and Sector Applications

Google announces that its technologies cover over 300 languages, which it associates with 7 billion people representing 86% of the global population. The company publishes interactive analyses via AI & Economy ATLAS to document global AI usage. In aviation, Google claims to have expanded its research to reduce climate impact, with applications reported in the UK in partnership with the government and in Asia. The company also mentions the aggregation of health, food security, and socio-economic data into a unique planetary engine, already mobilized during an Ebola outbreak in the Democratic Republic of the Congo and in the United States to identify vulnerable communities through 21 CDC indicators.

Education: Promises, Limits, and Teacher Training

Google notes that about 90% of children of primary school age are enrolled in school and that approximately 87% complete this cycle, while citing 2022 studies on inequalities in access, quality, and availability of multilingual content, and a decline in learning outcomes measured in 81 countries over the past two decades. The company asserts that AI can personalize learning, break down language barriers, and clarify difficult subjects using tools like Guided Learning, with free access to student tools in several countries, and support for teachers in research, preparation, and administrative tasks. However, it raises challenges regarding accuracy, security, critical thinking, cheating, and loss of learning. Google indicates that it has partnered with ISTE+ASCD to offer AI literacy training to 6 million teachers in the United States. The company also connects these efforts to supporting workers and reducing language barriers.

Positioning and Precautions: Collaboration and Scope of Advances

Google presents its actions as a use of AI to accelerate scientific advancements for the benefit of the population, citing contributions to early disease detection and natural disaster prediction. The company argues that AI advancements can accelerate science and directly improve lives, with a focused effort on health, risk forecasting, learning, and the economy. However, it acknowledges that the benefits are not guaranteed and require societal collaboration to realize gains and mitigate risks, while believing that current advancements allow for questions once deemed out of reach. In this context, models like WeatherNext 3 are presented with attention to historically underserved regions in terms of data, through approaches combining satellite observations and AI.