Forscher warnten, dass in beliebten Datenbanken zur Erfassung von Tierarten Hunderte gefälschte KI-Bilder entdeckt wurden. Naturfotografen verwenden häufig KI, um ein Bild zu bearbeiten und zu verbessern, aber der Algorithmus kann Teile verschiedener Arten einbeziehen, um das neue Bild zu erstellen.
https://www.theguardian.com/environment/2026/jul/20/ai-slop-manipulated-fake-images-birds-citizen-science-aoe
30 Kommentare
The rise of generative AI platforms such as [ChatGPT](https://www.theguardian.com/technology/chatgpt)and Google Gemini has led to a sharp increase in fake and enhanced images of rare and celebrated birds across wildlife photography forums. Users can create high-quality fake images in seconds – or enhance a photograph by asking the AI to remove a branch or leaf that might be obscuring part of the subject, which can inadvertently introduce significant changes.
In a [recent commentary](https://www.nature.com/articles/s41559-026-03141-y.epdf?sharing_token=JEu_0kUalhm-oSGzZpOc4NRgN0jAjWel9jnR3ZoTv0MIXYSGF7IH-YSv2oP4xX-JIwr_LTtxwIQ6LrDFLSyJ2_Eyv-ExAWdrUSm-v7qgMPul0_3OcbDpubaY965CMem4OU-QMBnLegaWuJA67Li19tYBnkdwnB0tpt71NKb_Nyc%3D) in the science journal Nature, researchers warned that hundreds of fake images have already been discovered on popular databases for recording species. They said that the true scale of the issue is unknown – as many could go undetected – and could contaminate records collected by the public.
“My experience of looking at Facebook these days is that a huge volume of wildlife photos now are simply AI-generated imagery,” said Dr Alexander Lees, an ecologist at Manchester Metropolitan University who authored the journal article. “The idea that we could maybe use those photos to help us understand where species are in space and time is very difficult.”
Outright hoaxes remain rare and are often easy to spot, said Lees – nobody is falling for a toucan sighting in Siberia – but he says birders often use AI to edit and improve an image, and the algorithm can then introduce parts from different bird species to create the new image.
Lees points to the example of a false sighting of a red-winged blackbird in central Brazil, which is usually found in North America and had never been seen in this area of Brazil until the reported sighting. The bird was, in fact, an epaulet oriole – a common [new world bird species](https://blog.mybirdbuddy.com/post/old-vs-new-world-birds) – but the photographer had asked an AI platform to make the picture “look better”, which is when parts of the red-winged blackbird were added, resulting in the false sighting.
“Wildlife photographers can be quite obsessed with getting a beautiful photo, but there’s a risk that the image might actually cause problems down the line when AI has been used to edit it,” said Lees.
https://www.nature.com/articles/s41559-026-03141-y
AI pollution is going to really screw up a lot of databases and information deposits I think, and we won’t notice them until they become needed.
There is also the small issue of the fact that birders can be incredibly competitive and people at times will literally make stuff up so that they can look better on their lists. Pictures used to be one way to guard against that so now we have AI to make that even harder.
There are of course simply well meaning people who just misidentify things but the above can be a serious problem. And some of the fighting over it can be quite the sight to behold.
There ought to be a standard set of allowed edits anyways, and AI outright banned
It’s been a problem on iNaturalist for a while, unfortunately.
I find it sad that we are experiencing the end of video and photo evidence meaning anything. Imagine how good all this software could be in 20 years at making effectively meaningless nonsense that is totally convincing compared to the real thing.
I’m sorry, but the moment you doctor an image said image should not be allowed in such a database. I get increasing brightness or contrast but you do that with a slider or in proper editing software, the old fashioned way. The nanosecond you unleash AI on a picture, the picture should be considered useless.
Even without AI, I’ve seen tons of terribly overedited images on eBird. This is probably somewhat inevitable when you try to entice people to your citizen science project by offering convenient tools for hobbyists—some people show up for the tools and not the science. People basically ignore media rating guidelines and a lot of photographers are clearly just treating it like a photo gallery to show off their portfolio.
Look up Baby Peacock and half the top results are AI slop.
Photographers using AI is an oxymoron. The photograph ceases to exist the second you use AI.
Why would you use AI to ruin your photo? Get better pictures
Hmmm the article says ‘1,400 of 6 million images on i naturalist appear to be Ai generated’ while the pull quote is about Facebook which is full of AI slop.
The thing about citizen science if it goes both ways – users on inat/ebird etc can rate and validate the images of others and ‘fake’ or anomalous sightings reported. I do think AI slop wildlife images are a problem but I don’t think they’ll ruin the citizen science databases
If the image has been processed with AI it has not been improved.
That’s concerning. Is there a source of truth similar to the National Archives? This is also becoming a problem for photos from WW2. Not just AI but photos of renactors with a filter can be very convincing. Essentially whenever I see a WW2 photo I didn’t recall seeing earlier I ask for the link to the national archives
Is the orange lipped monkey real?
This is happening to ALL datasets on the internet. It’s just enshitification and training itself on its own slop.
Ghost of Audubon himself…making up species here we are…
Well drat. Guess we can’t trust any news of new species anymore.
I remember when AI images were exclusively prismatic fractal eye-dog dreamslop.
it’s like copying something on a printer and it starts going askew each copy. Then parts of it start going off the page entirely.
Man I never even considered how AI is going to affect *history.*
People in the future are going to have to sort through AI to find anything real from previous generations.
AI is just making everything worse I don’t know why we delude ourselves into believing these techno leaching welfare queens who steal our water and taxes for billions are “helping”.
I tell friends to google for „baby peacock images“ if they tell me that google can absolutely be trusted for image search results.
Dead internet theory is no longer a theory — it’s our new reality.
Kind of reminds me of how low-background steel is super valuable because nuclear testing fucked up the atmospheric radiation levels so much. Nov 2022 is going to forever mark the date that the trustworthiness of information and evidence changed due to a new technological advancement, and sources of information from before that date is going to be regarded more highly.
We’re very quickly going to go from having all of human knowledge at our fingertips to having a bunch of untrustworthy material.
We are one step away from rapid devaluation of anything created after 2022. Old photos and old music will be gold (but only if reliably proven to be pre-2022). Ironic that this is also the time when we became more productive than ever. Something similar happened to music when high quality recording became available at home in the early 2000s. Everyone had expected that it would democratize music business and we would have so many more great artists and bands! In the actuality the opposite happened – too much noise, impossible for someone properly talented to break through all that. So it turned out that the hard road that people had to take to get some pro studio time before was actually doing the vetting.
Yeah the problem with using AI for photo editing is that it’s destructive. Unlike in something like photoshop where you can highlight or lasso or otherwise designate an area and alter it or remove it without touching the rest of the image, AI says, „sure I’ll remove that background for you“, but it has to recreate the entire image to do it. And it invariably introduces hallucinations and artefacts in the process. I’ve used AI for very low stakes personal masking jobs and noticed this behavior pretty frequently. It’s shown me that AI is still not ready for primetime when it comes to image editing. Still has a long way to go.
I’m not saying they *would* do it, but couldn’t all the image generators simply include some kind of metadata in the .png / etc. saying that it’s AI? (If not metadata… maybe a specific string, or something?)
If you could easily „vet“ a generated image, I think people would be able to calm down about AI, at least a bit. (And that’s good for business. ^I ^assume…
So AI stole and ingested real information and is now spewing out wrong information which people may take as real and which AI will then re-ingest as real resulting in even more distorted output. Is there a term in common use for this phenomena, or is „eating your own vomit“ the best we have.
It’s a problem. Even modern digital cameras with automatic post editing can introduce fabrications.