AI-enhanced bird photos contaminate citizen science databases
Researchers warn that generative AI tools like ChatGPT and Google Gemini are unintentionally contaminating citizen science databases used for biodiversity research. When users enhance bird photographs by removing branches or adjusting colours, AI algorithms can inadvertently introduce features from different species, creating false species records that undermine the credibility of platforms like iNaturalist and the Macaulay Library.
What happened
Users of generative AI tools such as ChatGPT and Google Gemini are uploading AI-enhanced bird photographs to citizen science platforms including iNaturalist and the Macaulay Library. When users ask AI to remove branches, improve lighting or sharpen images, the algorithms can unintentionally introduce features from other bird species, creating false species records.
Researchers documented a case where an AI-enhanced photograph of an epaulet oriole appeared as a red-winged blackbird in central Brazil, a species never previously recorded in that region, after the photographer used AI to improve the image.
Reported impact
Scientists warn that hundreds of fake images have already been discovered on popular databases for recording species, though the true scale is unknown as many could go undetected. The contaminated records threaten the credibility of citizen science platforms routinely used in scientific research to monitor species' habitat range, migration patterns and the effects of climate change.
Index analysis
Experts note that the issue stems from subtle, unintentional alterations rather than deliberate hoaxes. The AI algorithms introduce identifying features from different bird species when enhancing photographs, making it harder for scientists to distinguish genuine sightings from digitally manipulated ones.
Classification
Sources and evidence
This record was researched and written by the Index, based on the following original reporting.