As bear encounters and attacks reach record levels across Japan, technology is blurring reality with videos created with generative artificial intelligence, wreaking new havoc.
Fake clips of a woman driving a car and a scene in a cafe where customers are fed to a bear have racked up millions of views, a phenomenon with potentially dangerous consequences.
Experts warn that these convincing illusions, often designed to elicit emotional responses, risk distorting public understanding of real wildlife dangers and undermining essential conservation and management efforts.
As the country faces an increase in bear attacks, the challenge is not only how to protect ourselves from the animals themselves, but also how to protect ourselves from the flood of misinformation that poses an even greater threat.
When you search for “bear videos” on TikTok, an endless stream of short, AI-generated videos begins.
One popular video shows a woman brandishing a broom to scare away a bear trying to break into a building. It was uploaded in mid-October and had over 800,000 views within a month.
Other videos show a woman hand-feeding a bear cub and falsely claim there is an interactive bear cafe in Hokkaido, a popular spot for bear encounters.
All clips are believed to be generated by AI.
Many include logos for Sora, OpenAI's video generation model, and VoyagerX Inc.'s Vrew, which creates videos from text. In some cases, the poster himself or herself may add a label such as “AI Edited”.
However, a watermark does not guarantee that viewers will recognize the footage as fabricated. These videos also have comments like “Is this real?” Or react in a half-satisfied manner.
science says otherwise
This issue is not just about misleading and confusing the public. Fake bear videos are now being interpreted and actively used to reinforce unsubstantiated claims about wildlife behavior.
One of the clips created by Sora shows a bear standing next to a row of solar panels and knocking them down. In the comments, some users claim this proves that bears are invading the town because solar projects are destroying their habitat.
“There is almost no scientific basis for that claim,” says Shinsuke Koike, a professor of ecology at Tokyo University of Agriculture and Technology.
Koike clarified that there is no evidence that the bears' behavior changed after the solar cells were installed.
As their habitat expands, it is not surprising that there is some overlap with some of these facilities. However, Koike argues that the assumption that “solar panels were installed in long-established bear habitats, pushing the animals into human habitats” is completely wrong.
He warns that such misconceptions can distort the public's understanding of why bear encounters are on the rise and lead to misplaced criticism of governments and policies.
Koike believes that this could “stagnate long-term bear control efforts and have serious consequences.”
Misconceptions that it is safe to feed bears can even put people's lives at risk.
Find sources
Fake bear videos also pose widespread risks to society.
In November, a TikTok user posted what appeared to be a news article claiming that a 1-meter-tall bear had invaded a convenience store. The incident was framed as another alarming intrusion of wild animals into human space. The video also identified the location as Noshiro City, Akita Prefecture.
But city officials say that while bears are frequently sighted in the area, no such incidents have occurred.
A source said, “We don't know the creator's intentions, but if it was made as a prank, it would be a nuisance.''
An even bigger concern is what will happen if fake videos can no longer be flagged as such.
Isao Echizen, a professor at the National Institute of Informatics and an expert in generative AI, said it is becoming increasingly difficult to distinguish between real and fake footage as technology evolves.
At the moment, watermarks such as the Sora logo are showing some signs, but they can be easily blurred, cropped, or edited.
Echizen warns that technology to completely remove watermarks may soon be available.
“It will no longer be reliable to identify AI-generated content just by checking the watermark,” he says.
This makes user vigilance even more important to avoid being fooled when scrolling. Echizen said social media is driven by an “economy of attention,” in which posts are not rewarded for factual accuracy, but rather for increased views.
“You have to realize that what you see in your feed isn't necessarily the truth,” he says.
This creates and maintains the habit of verifying content with trusted primary sources.
(This article was written by Ryoka Tominaga and Mitsuteru Furuhata.)
