2025 was a rough year for human-bear relations. In the U.S., Florida had its first-ever fatal attack from a black bear, and two men died in Arkansas in separate incidents. Japan’s black bear population surged in 2025, resulting in so many fatal attacks on humans that the country’s military got involved.
And in perhaps the most terrifying example of all, a grizzly bear attacked a group of schoolchildren in British Columbia, Canada — and then disappeared beyond the reach of wildlife officials.
In many cases, these attacks prompt calls for renewed hunting of bears. That’s even true in British Columbia, which hasn’t allowed grizzly bear hunting since 2017. But state-sanctioned hunting doesn’t always reduce conflicts with bears, because bears often seek out cities for their abundant trash and other food sources.
Instead, several research programs have begun exploring how artificial intelligence (AI) can help prevent bear attacks. So far, this research includes multiple approaches. From facial recognition of “problematic” bears to high-tech radar warning systems, wildlife advocates hope the new methods can help keep both humans — and animals — safe.
“We’re trying to give people tools to prevent conflicts,” Elbert Bakker, a research support specialist for Polar Bears International, told GearJunkie. “There’s lots of potential for this technology.”
BearID Project
Facial recognition technology has been around for years. Facebook infamously rolled out facial recognition on its platform, and then backpedaled when users balked at the many problems it posed, from privacy concerns to false positives.
But bears likely don’t have an issue with it. That’s where the U.S.-based BearID Project comes in. The research initiative has developed facial recognition to help wildlife officials identify bears and improve wildlife management. It could also help identify bears involved in attacks.
It’s a potential game-changer because many bears lack markings to make them easily identifiable, even for wildlife biologists. The project combines AI-based facial recognition software with video from remote camera traps. This software tool could “aid conservation efforts worldwide,” the project’s lead scientists wrote.
The November attack in Bella Coola, British Columbia, is an example of how the technology could help. Canadian wildlife officials spent 3 weeks trying to find the grizzly bear that injured 11 people, most of them children. Though they found several grizzlies, officials were unable to definitively connect those animals to the Bella Coola attack, and eventually gave up the search.
Wildlife Tracking
But the BearID Project could also do more than identify bears involved in attacks. The technology can also improve tracking of individual bears, allowing wildlife managers to make better decisions for both humans and animals, according to The Vital Ground Foundation, a conservation group for grizzly bears.
“The BearID Project represents the potential for new technology to help bears and people coexist on our shared landscapes,” Matthew Hart, communications director for the group, told GearJunkie. “By providing noninvasive insights into the ever-fascinating lives of individual bears, it can help wildlife managers prevent conflicts before they happen and help conservation groups like us identify the most important movement areas to protect.”
Bear Radar in the Arctic
Right now, in the far reaches of the Canadian Arctic, the polar night never ends. From November to February, the sun remains below the horizon, leaving a constant darkness for the few humans who live in this frigid ecosystem.
But the handful of workers in Eureka — one of the most remote research stations in the Canadian Arctic — have an extra set of eyes peering into the darkness: a high-tech radar system being created by Spotter Global and tested by nonprofit group Polar Bears International.
Last summer, this Canadian nonprofit installed three radar units in Eureka. It’s a test to see if the technology, which they call “bear-dar,” can serve as an early-warning system to let workers know when a polar bear is approaching.
The system leverages advances in machine learning, a form of AI that uses algorithms to learn from data and improve over time at a specific task. In this case, the group’s researchers are hoping the radar system will learn how to recognize the movements of polar bears. It’s a difficult process, as the radar system must learn to differentiate between the movement of human workers, other Arctic animals like wolves and muskox, and environmental factors like crashing waves.
“We wanted to prove it in extreme conditions, which is why we’re in Eureka, which is pretty far north,” said Bakker. “If we can prove that it’s working here, it should be easier elsewhere in the Arctic.”
Sneaky Bears
You might think that spotting a polar bear in the treeless landscape of the Arctic would be easy, but you’d be wrong. Despite their huge size, polar bears are apex predators that rely on stealth and camouflage to catch their prey.
That’s what happened when a polar bear killed 34-year-old Christopher Best on Brevoort Island in August 2025. The island serves as another research station in the same archipelago as Eureka.
Best was outside taking photos of a distant polar bear when a second bear snuck up behind him, cutting off his escape route, CBC reported. Best’s mother told the CBC that his death could have been prevented with improved warning systems, which is exactly what the “bear-dar” is meant to do.
“If we can get [the radar] isolated to large animals moving toward camp, and give warnings of that, the people working nearby can just be doing what they do until they get a warning — rather than suddenly seeing one walking around a building,” said Bakker, a lead researcher for Polar Bears International.
Other Use Cases for Bear Detection
Both facial recognition and “bear-dar” could be useful outside of their initial testing of grizzlies in British Columbia or polar bears in the Arctic.
Two members of the BearID Project team, Mary Nguyen and Ed Miller, helped bring their research to Ecuador in 2024. They spent 4 weeks working with researchers at the Universidad San Francisco de Quito to “accelerate” their camera trap analysis with AI.
In this case, Miller and Nguyen helped the Ecuadoran school bring the AI-based video analysis to its research on Andean bears, also known as spectacled bears. They even began experimenting with using the technology to identify individuals of other native species, like pumas, Miller wrote in a LinkedIn post.
As for the early warning system of Polar Bears International, it’s still in the early stages. However, the group is already discussing further testing of the large polar bear population in Svalbard, Norway. It’s also possible the technology could be adapted for other bear species in other environments, according to Bakker, like campgrounds or certain areas of national parks.
More research is needed to determine where these technologies can be applied and how they might work in other environments. But as bear populations grow and attacks increase, these scientists hope technological advances can allow humans to co-exist with these beloved animals.
Read the full article here
