In Yellowstone National Park, a network of 48 autonomous recording units listens around the clock. The devices are designed to capture every howl, bark, and chorus that drifts across the landscape, feeding audio into machine-learning models trained to decode wolf communication in real time. It is, according to Yellowstone Wolf Project’s Dan Stahler, the most detailed acoustic study of wild wolves ever conducted. Behind it stands Colossal Biosciences and its nonprofit arm, the Colossal Foundation.
Best known for its de-extinction work, including the headline-grabbing return of the dire wolf, Colossal has spent the past year quietly building out a conservation technology portfolio that extends well beyond genetics. At its center is a suite of AI-powered bioacoustic tools capable of monitoring, identifying, and ultimately protecting species that are difficult or nearly impossible to observe by traditional means.
Listening to Yellowstone’s Wolves
The Yellowstone bioacoustics effort is a partnership between the Colossal Foundation, the Yellowstone Wolf Project, Grizzly Systems, and Yellowstone Forever. The team has verified more than 7,000 unique howling events, using that dataset to train AI classifiers that can now identify individual and chorus howls with over 92% accuracy. Four wolves wear audio-logger collars that pair sound with GPS and motion data, enabling researchers to link vocalizations to specific locations and behaviors without setting foot on the wolves’ habitat.
The custom-built models do more than count howls. They cluster calls by acoustic fingerprint, extract pack identity, estimate group size, and can even flag the presence of pups. When a gunshot registers on the recorders, the system flags that too. The practical result is a non-invasive, continuously running census that gives researchers a window into wolf pack dynamics they never previously had.
“By merging AI, acoustics, and rugged field hardware, we’re addressing some of the major challenges afflicting conservation and creating a scalable framework that can support coexistence wherever wolves roam,” said Ben Lamm, CEO and co-founder of Colossal Biosciences.
Rediscovering a Bird Lost for Over a Decade
The same platform that listens for wolves in Wyoming is now helping locate a bird that most scientists believed was gone from the South Pacific. The tooth-billed pigeon, known locally as the manumea and sometimes called the “little dodo” for its close relation to that extinct species, was last photographed in 2013. Fewer than 100 are estimated to survive.
In a collaboration with the Samoa Conservation Society, the Colossal Foundation deployed its bioacoustic classifier on the island and detected the pigeon’s calls 43 times during its first deployment.
What makes the achievement remarkable is how little data the AI required to get there. Colossal’s team trained its machine-learning algorithm on just five minutes of existing manumea audio recordings. From that slim foundation, the classifier achieved 95% accuracy in detecting the bird’s unique vocalizations. Colossal has since open-sourced the algorithm, making it available to other conservationists searching for lost or elusive bird species worldwide.
“Colossal’s AI capabilities have allowed us to successfully identify and recognize the tooth-billed pigeon’s distinctive vocalizations and we can now track a species that hasn’t been photographed in over 13 years,” said Matt James, Executive Director of the Colossal Foundation. “We’re excited to deploy additional monitoring systems in the field and see what this means for the future population of the species.”
A Scalable Tool for Conservation
The speed with which the technology adapts to new species is part of what sets it apart. When Colossal’s team was asked on a Friday whether the bird classifier could be retooled for wolves, senior AI scientist Abhishek Jana reportedly had a working wolf classifier ready by Monday morning. That kind of agility suggests a platform with broad applicability across ecosystems and taxa.
The bioacoustic work is part of a broader conservation technology push documented in the Colossal Foundation’s 2025 Impact Report, which spans more than 40 species across six continents. Alongside the acoustic monitoring work, the foundation has deployed thermal-equipped drones to track elephant movements in Africa and partnered with the Baylor College of Medicine to develop an mRNA vaccine for a fatal elephant disease. The foundation has deployed over 20 frontier technologies in its inaugural year, backed by $100 million in total funding.
For Colossal, the conservation applications of its AI tools are deeply intertwined with its de-extinction mission. Technologies developed to bring back species like the dire wolf create new capabilities that can be redirected toward the many species that still exist but are in peril. Sound, it turns out, is an especially powerful bridge between those two goals.
“The AI, acoustics and sound classifications we’re building at Colossal demonstrate the effectiveness of AI-powered bioacoustic technology in low-data scenarios,” Lamm said. “Our unique techniques have already helped in Samoa and will be applied to other endangered birds in other habitats.”
As the Colossal Foundation scales its monitoring networks from Wyoming to the South Pacific, the same question keeps surfacing: how many species thought to be silent or lost are still out there, waiting to be heard?
