Google is using AI to map endangered species before they disappear

Scientists estimate that nearly one million species face the threat of extinction. Losing them doesn't just wipe out wildlife; it tears away the biological foundations of our food systems, climate stability, and modern medicine.

Historically, mapping out a single species' genome was excruciatingly slow and brutally expensive. Now, Google's AI tools are stepping in to turn a multi-year slog into a matter of days.

Cracking animal DNA in days

The Vertebrate Genomes Project, spearheaded by Erich Jarvis at The Rockefeller University alongside the Earth BioGenome Project, is using Google's machine learning pipelines to sequence life on Earth.

Google.org recently awarded The Rockefeller University funding under its AI for Science initiative to sequence 150 additional endangered species, including Grevy’s zebras and African penguins. Every bit of that genomic data will be open and free to researchers.

This follows an initial run preserving the genetic sequences of 13 endangered species across mammals, birds, reptiles, and amphibians, with genomes for nine already published.

To put the speedup in perspective: sequencing the very first human genome took 13 years and $3 billion. Today, Google's AI tools—namely DeepVariant, DeepPolisher, and DeepConsensus—help scientists churn out stellar-accuracy genomes for humans, animals, and plants in mere days for just a few thousand bucks.

This isn't just academic cataloging. When researchers at the University of Otago mapped the DNA of every living kākāpō—the flightless, nocturnal parrot—they used the data to run a targeted breeding program that pulled the bird back from the edge of oblivion.

Why it matters

We think this is one of the sharpest applications of Google's specialized AI stack. Instead of burning compute purely on chat agents, applying tools like DeepPolisher to messy biological readouts solves an actual bottleneck in wet-lab genomics.

For conservationists, the barrier to entry has essentially collapsed. Projects that previously required national consortiums and millions in grant money can now be carried out by smaller university teams to guide real-time breeding and disease prevention.

There is, however, an uncomfortable catch. AI can digitize biological diversity at breakneck speed, but storing high-fidelity genetic data in a cloud database doesn't stop habitat destruction or poachers. If governments treat open-source genomic sequencing as a substitute for on-the-ground preservation, these models risk becoming little more than high-tech biological obituaries.