BOULDER, Colo., July 21 — Hundreds of fake AI-generated images have infiltrated the world’s leading biodiversity databases, a contamination campaign that threatens the scientific foundations of conservation planning and species protection. Researchers from the University of Colorado Boulder and the University of Oxford have documented the problem across platforms including iNaturalist, eBird, and the Global Biodiversity Information Facility (GBIF), warning that without proactive measures the integrity of a generation of ecological data could be compromised.
The scale of the data at risk is staggering. iNaturalist, launched in 2008 and now a joint project of the California Academy of Sciences and the National Geographic Society, hosts over 100 million observations from more than 3 million users as of 2023. eBird, started by the Cornell Lab of Ornithology in 2002, contains over 1 billion bird sightings contributed by 600,000 registered users. GBIF, an international government-funded network established in 2001, aggregates over 2 billion species occurrence records from 1,700 institutions.
These three platforms alone constitute the backbone of modern citizen-science biodiversity monitoring.
How AI-Generated Images Are Polluting Scientific Databases
The problem came to light in 2023, when platform moderators spotted images with impossible features — extra limbs, fused body parts, anatomical chimeras that could only have been produced by generative AI models. These images, often created by wildlife photographers using editing tools powered by generative adversarial networks (GANs) and diffusion models, can inadvertently merge traits from multiple species when asked to “improve” a photo. What emerges is a species that never existed, but one that now occupies a slot in a database trusted by scientists worldwide. iNaturalist acted quickly.
By 2024, the platform had publicly banned AI-generated content and deployed automated detection tools to flag suspect submissions. GBIF updated its data quality guidelines in 2024, creating a framework to identify and isolate records that may be based on synthetic imagery.
But the detection arms race is only beginning. As AI image generation becomes cheaper and more accessible, the volume of contaminated submissions is expected to rise. Strategically, the calculus for conservation groups is shifting.
The International Union for Conservation of Nature (IUCN), founded in 1964, relies on citizen science data for its Red List of Threatened Species — the most authoritative global assessment of extinction risk. Misidentified images can lead to false species records, skewing range maps and wasting scarce conservation resources on phantom populations or misdirected habitat protections.
A single fake observation in the wrong database could ripple through a Red List assessment, altering a species’ status and diverting funding.
The Broader Stakes for Evidence-Based Conservation
What this signals is a new vulnerability in the digital infrastructure that underpins environmental policy. Democracies have invested heavily in open-data platforms that invite public participation; those same platforms are now exposed to a form of data contamination that is difficult to detect at scale. The researchers from Boulder and Oxford have warned that without coordinated international standards and automated verification systems, the problem will accelerate as generative models improve faster than the tools to police them.
The broader stakes extend beyond any single database. Species distribution models, conservation planning exercises, and government environmental assessments all draw on GBIF and iNaturalist records.
If the underlying data cannot be trusted, the policy derived from it rests on uncertain ground. For the United States and its allies, which rely on such data to meet international biodiversity commitments, the contamination represents a direct challenge to evidence-based governance. In the months ahead, platform operators face a strategic choice: invest aggressively in detection technology and human moderation, or watch the value of their datasets erode.
The IUCN and other institutions that depend on citizen science will need to recalibrate how they vet observations. The fake images are not an isolated incident — they are the opening move in a longer contest between scientific integrity and the ease of AI-generated deception.
How the biodiversity community responds will determine whether the world’s most comprehensive picture of life on Earth remains clear, or whether it becomes an unreliable mosaic of machines and nature.
Sources
{“@context”: “https://schema.org”, “@type”: “NewsArticle”, “headline”: “Fake AI Images Disrupt Global Biodiversity Databases”, “description”: “BOULDER, Colo., July 21 — Hundreds of fake AI-generated images have infiltrated the world’s leading biodiversity databases, a contamination campaign that threatens the scientific foundations…”, “datePublished”: “2026-07-21T09:30:00+00:00”, “author”: {“@type”: “Person”, “name”: “Noah King”}, “publisher”: {“@type”: “Organization”, “name”: “InfoPulse Today”, “url”: “https://newsfeed.infopulsetoday.com/”}, “citation”: [{“@type”: “CreativeWork”, “url”: “https://en.wikipedia.org/wiki/Artificial_intelligence”}]}
