The scientific community has encountered a new threat stemming from artificial intelligence. Researchers from the UK have found that images of animals generated or even simply edited using neural networks can significantly distort data in global observation databases. The problem lies in the fact that even minimal image enhancements, such as sharpening or color correction, can alter key species traits—beak shape, fur color, or wing patterns.
False Records and Errors in Ecology
Such manipulations create a risk of falsely recording an animal in a new region, misleading ecologists and biologists. For example, a slightly altered image of a bird might be mistakenly identified as a rare migratory species, affecting habitat range statistics and conservation decisions. This poses a particular danger to citizen science platforms, where volunteers upload thousands of images daily without strict verification.
A Call for Labeling and Verification
The study's authors urge administrators of such platforms to implement mandatory labeling of images that have undergone AI processing. Algorithms for automatic detection of traces of generation or editing need to be developed. Without this, data collection systems risk losing reliability, and scientific conclusions based on such databases could become erroneous.
Expert Commentary: This is an alarming signal for the entire data collection ecosystem. In a world where AI is already capable of faking faces and voices, the scientific community must act proactively. The introduction of cryptographic tags and blockchain verification for images could provide a reliable solution, ensuring the authenticity of materials and protection against unintentional distortions.