Google says its AI now touches 300 languages and 7 billion people. Here's what that actually means.
From spotting missed breast cancers to warning Jamaicans about a hurricane, Google is rolling out AI tools that reach far beyond the chatbot.
Key points
- Google announced its products now support more than 300 languages, covering about 7 billion people, or 86% of the world.
- A breast cancer study with Imperial College London and the UK's NHS found AI spotted 25% of interval cancers previously missed across mammograms of 175,000 women.
- AlphaFold, Google DeepMind's protein-structure tool, has predicted all 200 million known protein shapes and is now used by 4 million researchers in 190 countries.
- WeatherNext 3, Google's new weather model, delivers precipitation forecasts a day ahead that are 50% more accurate.
- Google's Flood Hub now covers 2 billion people across more than 150 countries.
Google used a company blog post this week to lay out, in one place, what its AI is quietly doing outside the chatbot window. A lot, it turns out: in hospitals and in the path of storms.
The headline number is language. Google says its products now handle more than 300 languages, reaching around 7 billion people, roughly 86% of everyone alive. For a nurse in Nairobi or a shopkeeper in Jakarta, it means search, translation and voice tools finally speak the language they actually use at home.
What is Google actually doing with AI in health?
The most striking piece is cancer screening. A study run with Imperial College London and Britain's National Health Service found an AI model caught 25% of so-called interval breast cancers, the tumours that slip past a first mammogram and turn up between routine scans, in a review of 175,000 women. Cancers that would've been missed, spotted earlier.
Google's diabetic retinopathy tool, which scans photos of the back of the eye for a common cause of blindness in people with diabetes, has now supported more than 1.15 million screenings worldwide, with a target of 6 million over the next decade.
On tuberculosis, where roughly 40% of infected people are never diagnosed, a chest X-ray model used by a partner called Nexus Intelligence has read over 25,000 scans across 40 sites in six countries. There's even a system that listens to coughs to flag likely TB cases.
Underneath all of this sits AlphaFold, the protein-mapping system from Google DeepMind that won a share of the 2024 Nobel Prize in Chemistry. It's predicted the shapes of all 200 million proteins known to science and is used by 4 million researchers across 190 countries, from drug hunters to teams studying neglected illnesses like Chagas disease.
Can AI really predict disasters?
Better than it could a year ago, yes. Google points to Jamaica, where officials used its WeatherNext model to track Hurricane Melissa's path accurately enough to secure early disaster funding and begin evacuations before landfall. Its Earthquake Alert system also pinged millions of phones in Venezuela ahead of a recent quake.
We've been following WeatherNext closely: our 10 September story found that the model predicts cyclone paths three days out with the same accuracy older systems only managed at two days, and the 8 September piece covered its satellite upgrade that cuts forecast costs sharply.
The new WeatherNext 3 model pushes precipitation forecasts a day or more out to 50% greater accuracy, and it runs without the huge supercomputers older weather models need. That matters most in poorer regions that never had high-resolution forecasts.
| Tool | What it does | Reach |
|---|---|---|
| Flood Hub | Riverine and flash flood forecasts | 2 billion people, 150+ countries |
| Monsoon forecasts | Rainfall guidance for farmers | 38 million farmers in India, 2025 |
| Wildfire alerts on Search | Timely fire information | 75 million users, 520+ alerts in 2025 |
| WeatherNext 3 | Hourly global weather model | 50% more accurate rain forecasts |
My honest read: the language milestone is the headline Google wants, but the quieter story is that these health and weather tools have moved from demo videos to real deployments with named hospitals and governments attached. Watch whether the scale numbers hold up once independent researchers get access.
One plain caution. Health screening tools like these run on medical images and, in some cases, sounds of your cough or voice. If a clinic offers you an AI-assisted screening, it's fair to ask who sees the data, where it's stored, and whether you can opt out. Useful isn't the same as private.
Pricing isn't really the point here. Most of these systems reach hospitals, governments and researchers through partnerships rather than consumer apps. The tools you can touch today, Search alerts, Translate, Flood Hub notifications, are free.
