'Distributed Denial of Science': Indirect Data Poisoning Could Industrialize Scientific Fraud

arXiv·medium signal

As AI increasingly automates scientific research, the authors ask whether a remote adversary can weaponize the honest use of AI in science, and empirically evaluate a new attack: indirect data poisoning that seeds corrupted evidence an AI research system will ingest and propagate. Historically fraud required a company's resources — deep pockets, ghostwriters, corrupt academics — but this attack lowers the cost to remote, scalable manipulation of scientific conclusions. It's an early warning for anyone building AI-for-research pipelines that trust upstream data.