Risks of abuse of large language models, like ChatGPT, in scientific publishing: Authorship, predatory publishing, and paper mills

Abstract

Key points Academia is already witnessing the abuse of authorship in papers with text generated by large language models (LLMs) such as ChatGPT. LLM‐generated text is testing the limits of publishing ethics as we traditionally know it. We alert the community to imminent risks of LLM technologies, like ChatGPT, for amplifying the predatory publishing ‘industry’. The abuse of ChatGPT for the paper mill industry cannot be over‐emphasized. Detection of LLM‐generated text is the responsibility of editors and journals/publishers.

Discontinuity between journals' registration and submission requirements for Open Researcher and Contributor ID (ORCID)

“…Those editors also need to reflect on what a mandatory ORCID for only the submitting/corresponding author might achieve, especially if some authors might be guest authors, for example, where authorships have been purchased via paper mills, or if authors place their name to papers whose content might have been generated by large language models, like ChatGPT. 30 Editors of nursing journals should consider that if they mandate ORCID from authors, then they, too, should add their ORCID, ensuring that all editors display their ORCIDs on the journal's editorial board page, and that those ORCID accounts are open to the public, complete, accurate and up to date, to ensure equitable two-way trust and accountability. Finally, to not create an impression of entrapment or deception related to ORCID, journals should ensure compatibility of information related to this identifier in both registration and submission stages, ensuring congruity with the journal's IFA.…”

“…Finally, editors of nursing journals need to reflect on whether implementing a mandatory ORCID for submitting authors is fair on authors who may wish to submit to their journals, but might be excluded from doing so, and if such an editorial policy infringes on their rights and freedoms. Those editors also need to reflect on what a mandatory ORCID for only the submitting/corresponding author might achieve, especially if some authors might be guest authors, for example, where authorships have been purchased via paper mills, or if authors place their name to papers whose content might have been generated by large language models, like ChatGPT 30 . Editors of nursing journals should consider that if they mandate ORCID from authors, then they, too, should add their ORCID, ensuring that all editors display their ORCIDs on the journal's editorial board page, and that those ORCID accounts are open to the public, complete, accurate and up to date, to ensure equitable two‐way trust and accountability.…”