Scope
This policy applies to authors, reviewers and editors of Saltus. It covers generative artificial intelligence tools (for example language models that write, translate, summarize or rewrite text, and tools that generate or alter images, tables or code). It is based on ICMJE recommendations, COPE guidance and the World Association of Medical Editors (WAME) position on AI use. It may be updated as those guidelines change.
AI cannot be an author
An AI tool does not meet ICMJE authorship criteria: it cannot take responsibility for the work, declare conflicts of interest, give final approval or hold copyright. For that reason it is not accepted as an author or co-author and must not appear in the author list or in the CRediT contributions. Human authors alone are responsible for the whole content of the manuscript, its accuracy, its references, and for ensuring there is no plagiarism or fabricated material.
Use by authors
Authors may use AI tools to improve language, grammar or translation, always under their own supervision and verification. They must disclose any use of generative AI in the manuscript (in Methods or an acknowledgements section, as appropriate) and in the declaration field of the submission form, stating: the tool and its version, the date of use, what it was used for and who reviewed the output. If none was used, they must write “Not used”. This statement is not needed for basic spelling or grammar checkers that only correct errors without generating content.
Data, images and scientific content
AI must not be used to fabricate, falsify or alter data, results or images (including clinical, radiological or histological images), or to generate references or citations. Figures created entirely with generative AI and presented as data are not accepted. Conceptual illustrations generated with AI are accepted only if clearly labelled as such with the tool declared, and they must not depict real patients. Authors must check every reference generated or suggested by an AI tool, because these tools can invent citations that look real.
Patient data and confidentiality
Data that identify patients, non-anonymized clinical information, or unpublished or confidential third-party material must not be entered into external AI tools. Anyone using these tools is responsible for respecting informed consent and data protection, especially in work involving minors.
Reviewers
Manuscripts under review are confidential. Reviewers must not upload the manuscript, any part of it, or review text that reproduces it, to external generative AI tools, because this may breach confidentiality and the authors' rights. The assessment must be the reviewer's own judgement: AI does not replace expert judgement. If a reviewer used a tool to improve the wording of their own report, they must tell the editor.
Editors and editorial staff
Editors follow the same confidentiality rules: they do not upload manuscripts or reports to external AI tools. No editorial decision (accept, reject or request changes) is made by an AI tool. If the journal uses AI-based programs for support tasks, such as similarity screening or format checks, it will use tools that protect confidentiality, and a person always reviews the result.
Non-compliance
Failing to disclose AI use, or using it against this policy, is handled as a research integrity matter following COPE flowcharts. Depending on severity, it may lead to requests for clarification, rejection of the manuscript, correction or retraction of an already published article, and notification to the authors' institutions.
Model statement
Example: “During the preparation of this work the authors used [tool name, version] on [date] for [purpose: for example, improving English wording]. After using it, the authors reviewed and edited the content and take full responsibility for the manuscript.” If no AI was used: “No generative artificial intelligence was used in the preparation of this manuscript.”