Scientists Warn 610 Million iNaturalist Images Face AI Contamination Risk
Updated
Updated · The Guardian · Jul 20
Scientists Warn 610 Million iNaturalist Images Face AI Contamination Risk
2 articles · Updated · The Guardian · Jul 20
Summary
Hundreds of fake wildlife images have already been found on citizen-science databases, prompting researchers to urge birders to limit AI editing that can distort species records used in scientific studies.
Generative AI tools can fabricate rare birds outright or subtly alter real photos—removing branches, for example, while adding traits from other species and creating false sightings.
One Brazil record of a red-winged blackbird turned out to be a common epaulet oriole after AI “improvement” inserted features from the North American species, illustrating how misidentifications can spread.
iNaturalist has flagged about 1,400 images for AI use out of more than 610 million, but researchers say the true scale is unknown because many altered images may go undetected.
Those records help scientists track range shifts, flowering times and behavior as the climate warms, so researchers say even non-malicious edits could weaken a real-time conservation data source.
AI is corrupting wildlife data. Can this same technology become conservation's greatest ally?
While AI fakes threaten wildlife data, could AI's energy use be a greater environmental threat?
Safeguarding Biodiversity Data: Addressing the Surge of AI-Generated Content in Citizen Science Platforms
Overview
The rapid rise of AI-generated and enhanced images is creating a major challenge for citizen science platforms like iNaturalist and eBird. As these platforms face a surge in fake nature photos, authentic observations risk being wrongly dismissed, while fabricated content can gain false credibility. This undermines trust in scientific imagery and threatens the integrity of vast datasets that are crucial for biodiversity research. The urgency of this issue is highlighted by real cases of AI spoofing, showing that both the scale and impact of AI contamination are growing quickly and require immediate attention to protect data reliability.