AI‑generated fraud warning
- Cardiologist Eric Topol said on April 23 that a paper in the Journal of Digital Health Implementation was “fraudulent” and AI-generated after it listed him as an author without consent. - Retraction Watch reported the paper was dated March 29, named six authors tied to major institutions, and the journal and publisher website went dark within hours of the complaint. - The case landed as Nature reported tens of thousands of 2025 papers may contain invalid AI-made references, widening concern over scientific fraud. (nature.com)
A research paper can look real without being real. Eric Topol said this week that one such paper used his name without consent and was generated by artificial intelligence. (retractionwatch.com) (substack.com) Topol, executive vice president of Scripps Research, said on April 23 that the article appeared in the Journal of Digital Health Implementation and that he “had nothing to do with it.” Retraction Watch reported the paper was dated March 29. (substack.com) (retractionwatch.com) The paper’s title was “Implementation Science for AI Integration in Digital Health Systems.” Retraction Watch said several listed authors were affiliated with prominent institutions and that the journal and publisher website went offline within hours of the complaints. (retractionwatch.com) The underlying problem is simple: text generators can assemble a paper-shaped document that reads like scholarship even when the authorship, references, or underlying work are false. The Committee on Publication Ethics says editors are already dealing with AI-assisted fake papers, paper mills, and manipulated peer review. (publicationethics.org) That risk is no longer hypothetical. Nature reported on April 1 that tens of thousands of publications from 2025 might include invalid references generated by artificial intelligence. (nature.com) Researchers are also building tools to catch the pattern. Binghamton University said this week that Ahmed Abdeen Hamed’s xFakeSci system can detect up to 94% of bogus papers, outperforming several common data-mining methods. (binghamton.edu) (nature.com) In the 2024 Scientific Reports paper behind that work, Hamed and Xindong Wu tested fake and real abstracts on cancer, depression, and Alzheimer’s disease. Their model posted F1 scores between 80% and 94%, versus 38% to 52% for baseline methods. (nature.com) Publication ethics groups are pushing process changes as well as detectors. The Committee on Publication Ethics says publishers need stronger checks on authorship, originality, and whether artificial intelligence is being used to write or review papers. (publicationethics.org) Topol’s complaint turned an abstract integrity problem into a personal one. A fake paper did not just invent findings; it appears to have borrowed real names to make the fraud look publishable. (substack.com) (retractionwatch.com)
