In a world where every photograph on a screen may conceal traces of poaching, the Chinese project 'Guardian of Endangered Species' offers an unexpected tool: artificial intelligence capable of recognizing ivory products, pangolin scales or antelope horns in a matter of seconds.
The illegal wildlife trade annually brings criminals up to 20 billion dollars, and a significant portion of the turnover has moved online. On e-commerce platforms and social networks, photographs of bracelets, powders and souvenirs blend in with ordinary goods, complicating the work of law enforcement.
The new tool, which received an award at the World Internet Forum in Wuzhen, uploads a snapshot and within a few seconds produces a result with three levels of confidence. At the maximum score, the system transmits the data to specialists for verification and subsequent action.
At its core lies a combination of Baidu's PaddlePaddle deep learning platform and the Wenxin model, supplemented by data from the International Fund for Animal Welfare. Since 2020, the project has gone through three iterations and now recognizes 34 categories of endangered species, including live animals and products made from them, with an accuracy of about 86 percent.
An important step is open access to the model. Any researcher, conservation organization or government agency in the world can freely use and refine the algorithm, turning the Chinese development into a shared resource for the global fight against poaching.
By April 2026, the system had analyzed half a million images, identified more than 50 thousand suspicious ones and contributed to the removal of over 18 thousand illegal listings. Such figures show how technology scales efforts that previously depended on manual review.
When algorithms become allies in protecting biodiversity, the planet has a real chance to preserve species whose disappearance irreversibly changes entire ecosystems.
