AI-Generated Fake Images Pose Threat to Biomedical Research Integrity

Researchers raise red flag about AI-generated fake images in biomedical research

The authors of an editorial published in the American Journal of Hematology, claim that "generative Artificial Intelligence can be exploited to produce fraudulent scientific images, either from scratch or by modifying existing visual materials to increase the realism of the final fabricated product." Authors Enrico M. Bucci, Professor of Biology at the Sbarro Institute for Cancer Research at the College of Science and Technology, Temple University, and Angelo Parini, University of Toulouse, highlight a growing concern in science: the use of artificial intelligence (AI) to create fake images that look like real research data. The paper, titled "The Synthetic Image Crisis in Science," explains how tools powered by AI are being used to make realistic but completely fake scientific images. These images may be capable of evading detection because they are not edited versions of real photos. Instead, they are generated from scratch using AI tools, which can avoid discovery because they don't include telltale features to distinguish them from real ones. "These tools are now usable by anyone, regardless of scientific training. Prompted correctly, they can simulate a study's entire visual apparatus in minutes," the authors say. Modern AI systems can create fake images based on simple text descriptions. For example, a user can ask for a Western blot showing a certain protein in treated cells, and the AI will generate a believable image, even though no experiment was ever done. The same tools can also be used to make subtle changes to real scientific images. These changes can include adjusting colors, moving parts of the image, or adding features -- all without leaving the clues that normal editing tools would. The tools used for image generators are trained using real scientific images and are now widely available to the public. As a result, peer reviewers and journal editors are starting to find synthetic images in submitted research papers. "It is crucial that the scientific community and the peer-review system adapt to this looming threat of fakery in scientific data," says Antonio Giordano, M.D., Ph.D., Professor at Temple University, Founder and Director of the Sbarro Health Research Organization (SHRO). "The concerns outlined by Bucci and Parini require updated protocols for things like documentation, transparency, and accountability in response to the new reality of a world with AI."

Researchers Raise Red Flag about AI-Generated Fake Images in Biomedical Research | Newswise

Newswise -- "Generative Artificial Intelligence can be exploited to produce fraudulent scientific images," say the authors of an editorial published in the American Journal of Hematology (AJH), "either from scratch or by modifying existing visual materials to increase the realism of the final fabricated product." The authors Enrico M. Bucci, Professor of Biology at the Sbarro Institute for Cancer Research at the College of Science and Technology, Temple University, and Angelo Parini, University of Toulouse, highlight a growing concern in science: the use of artificial intelligence (AI) to create fake images that look like real research data. The paper, titled The Synthetic Image Crisis in Science, explains how tools powered by AI are being used to make realistic but completely fake scientific images. These images may be capable of evading detection because they are not edited versions of real photos. Instead, they are generated from scratch using AI tools, which can avoid discovery because they don't include telltale features to distinguish them from real ones. "These tools are now usable by anyone, regardless of scientific training. Prompted correctly, they can simulate a study's entire visual apparatus in minutes," the authors say. Modern AI systems can create fake images based on simple text descriptions. For example, a user can ask for a Western blot showing a certain protein in treated cells, and the AI will generate a believable image, even though no experiment was ever done. The same tools can also be used to make subtle changes to real scientific images. These changes can include adjusting colors, moving parts of the image, or adding features -- all without leaving the clues that normal editing tools would. The tools used for image generators are trained using real scientific images and are now widely available to the public. As a result, peer reviewers and journal editors are starting to find synthetic images in submitted research papers. "It is crucial that the scientific community and the peer-review system adapt to this looming threat of fakery in scientific data," says Antonio Giordano, M.D., Ph.D., Professor at Temple University, Founder and Director of the Sbarro Health Research Organization (SHRO). "The concerns outlined by Bucci and Parini require updated protocol for things like documentation, transparency, and accountability in response to the new reality of a world with AI." The authors warn that AI-generated images could damage trust in science if not addressed. It calls for new methods to detect these images and protect the quality of scientific research. Funding Acknowledgement: This research was funded by the Agence Nationale de la Recherche (ANR-23-IAHU-0011). About Sbarro Health Research Organization (SHRO) The Sbarro Health Research Organization conducts groundbreaking research in cancer, diabetes, and cardiovascular disease. Based in Philadelphia, Pennsylvania, on the campus of Temple University, SHRO's programs train young scientists from around the globe, accelerating the pace of health research and innovation.