Medical journals, law reviews, and literary magazines…
The New England Journal of Medicine and other academic publications are struggling to detect AI-generated submissions and manipulated imagery in research
- Incident date
- May 2024
- Target
- New England Journal of Medicine
Scientific and academic publications, including the New England Journal of Medicine, are grappling with the rising prevalence of AI-generated content in submissions. Editors are increasingly challenged to distinguish between human-authored work and machine-generated prose, forcing a reevaluation of editorial policies and verification processes.
What happened
The integration of AI into academic publishing has created significant operational hurdles. Editors at MIT Press have reported a notable increase in submissions suspected of being AI-generated, often lacking required disclosures. These submissions can strain resources, as platforms used to manage the influx of papers charge fees ranging from $10 to $32 per submission. Common indicators of AI involvement include specific repetitive syntax, heavy reliance on em-dashes, and formulaic phrasing.
Beyond text, AI-driven image manipulation has emerged as a critical concern for scientific integrity. In May 2024, the New England Journal of Medicine issued a retraction for a case study concerning "bronchial casts." The authors admitted to using AI to manipulate an image, specifically to straighten a crooked ruler featured in the visual data. Dr. Eric Rubin, editor-in-chief of the journal, described the incident as a failure of oversight and noted that the publication is currently experimenting with AI detection tools to identify machine-generated opinion pieces, though their efficacy remains uncertain.
While some editors acknowledge that AI tools can be beneficial for non-native English speakers to improve communication of authentic data or for fact-checking large statistical datasets, the threat of fabricated science remains a primary concern. Academic institutions are responding by implementing transparency requirements, such as those adopted by the Yale Law Journal and MIT Press, which mandate disclosure of AI usage in research and writing. Despite these measures, the rapid evolution of generative technology means that editorial teams are frequently forced to adapt their verification methods in reaction to new instances of AI-assisted misconduct.