Modern algorithms for generating and modifying images, including those based on neural networks, pose a serious threat to scientific databases, especially in the fields of biology and ecology. A research group has found that even routine enhancement of animal photos—such as adjusting lighting or contrast—can alter key species traits, like fur color or fin shape. This can lead to false identification and the recording of an animal in a region where it has never lived.

Citizen science platforms, where volunteers upload and classify images, are particularly at risk. Without rigorous verification, such platforms risk being filled with distorted data, undermining trust in scientific conclusions. The authors strongly recommend implementing mandatory labeling of AI-processed images and strengthening verification mechanisms—for example, using blockchain solutions to track the chain of changes for each photo.

This mirrors the situation with cryptocurrencies, where fake transactions or counterfeit wallets can collapse trust in an entire network. Scientific databases are a kind of "registry of truth," and protecting them from digital manipulation requires no less care than blockchain security. Until the community develops verification standards, every uploaded photo could become a time bomb for ecological research.

Expert opinion: As an analyst, I see parallels between this problem and risks in the crypto industry: in both cases, the cost of error is high. Embedding cryptographic metadata into images could be a solution, but for now, the scientific community is only beginning to grasp the scale of the threat. I recommend that citizen science platforms immediately audit their databases for AI-induced distortions.