Transparent AI, trustworthy publishing
BMJ Group introduced guidance on the responsible use of generative AI in scholarly publishing in 2023. By October that year, only 24 of the world’s 100 largest academic publishers had issued guidance on its use. 6,7,8,9 Our AI use policy 10 requires authors to disclose whether and where AI was used and to remain accountable for all submitted work; editors may reject content where its use is judged inappropriate or has not been disclosed. Findings from BMJ Group’s subsequent disclosure study are now informing revisions to the policy, submission questions and guidance for editors. Evidence informing policy Mandatory AI disclosure questions introduced across 49 BMJ Group journals generated data from more than 25,000 research submissions. The resulting analysis, published in JAMA, 11 is now informing the next iteration of BMJ Group’s AI use policy and the submission process used to collect, assess, and report AI disclosures. The analysis showed that 5.7% of authors disclosed AI use, compared with survey estimates of 28% to 76%. Although not directly comparable, the findings suggest possible underreporting or uncertainty about disclosure requirements, highlighting the need for clearer guidance. 11
BMJ Group AI policy recognised as leading practice An independent analysis of guidance from the world’s 100 largest academic publishers identified BMJ as one of three examples of “well crafted guidelines”. The researchers highlighted the policy’s clear disclosure requirements and its insistence that authors remain accountable for work produced using generative AI.
Examples of journals with well crafted guidelines included BMJ, Physiological Reviews, and PLOS ONE.” 9
8
Powered by FlippingBook