The systemic breach of AI in academic publishing: Infrastructural failures and authorial misconduct across seven major publishers

In academic publishing, there is a crisis that is both authorial and infrastructural; however, the discussion around the topic is heavily skewed toward the former rather than the latter. This paper presents an empirical map of both issues. Using a convenience sample of N = 323 retracted articles obtained from the Retraction Watch Database from seven of the major publishers: Springer Nature, Elsevier, Wiley, SAGE, Taylor & Francis, Emerald, and MDPI, we have used a mixed methods approach, namely, qualitative content analysis of the retraction notices and quantitative reporting of the coded frequencies. We coded the notices along the dimension of infrastructural vs. authorial failure. Our results show that even though algorithmic forgery (“AIgiarism”) does represent a real and growing threat, it is dwarfed in our retraction sample by the infrastructural issues in terms of scale, such as rogue editors, syndicated paper mills, broken peer review systems, and fake human-based clinical trials. This becomes clear when examining these phenomena publisher by publisher, revealing different infrastructural vulnerabilities that cannot be fixed regardless of how rigorously the policies are enforced by individual editorial offices.