Artificial Intellegence Policy

Purpose and scope

This policy sets out acceptable uses of artificial intelligence (AI), including large language models (LLMs), for manuscript creation, peer review, and editorial workflows. It applies to all submission types and all participants in the publication process.

Core principles

  • Human accountability: People (not tools) are responsible for the accuracy, integrity, and legality of all content and decisions.
  • Transparency: All material AI assistance must be disclosed.
  • Confidentiality: Unpublished manuscripts and reviews must not be exposed to tools that may store or train on the content.
  • Compliance: Follow relevant funder, institutional, and legal requirements in addition to this policy.

 

Authors

Permitted with disclosure

  • Language editing, formatting, grammar, and clarity.
  • Summarising prior work, generating outlines or alternatives for phrasing.
  • Code assistance for non-novel scripts, and figure or image editing that does not fabricate or misrepresent data.

Prohibited

  • Generating novel scientific content (e.g., data, results, analyses, references) that is not independently verified.
  • Fabricating, manipulating, or enhancing data, images, or audio beyond accepted scholarly standards.
  • Listing AI tools as authors or co-authors.
  • Uploading confidential data, proprietary materials, patient/participant information, or embargoed content to public AI systems that may log or train on inputs.

Requirements

  • Disclosure statement in the manuscript’s Acknowledgements or Methods. Authors must verify all outputs and accept responsibility.
  • Citation: If an AI system meaningfully informs content, cite it as software (with version/URL or DOI) in References where appropriate.
  • Data and code: Analyses assisted by AI must be fully reproducible without the tool (provide code, parameters, seeds, and documentation).
  • Images: For AI-generated or AI-altered images, clearly label as such in figure captions and comply with field-specific image-integrity standards.
  • Authorship: AI cannot meet authorship criteria. Human contributors must meet authorship standards and approve the final manuscript.

 

Reviewers

Permitted with disclosure to editors (not to authors)

  • Using local or enterprise (not for public or used for training) AI tools for language polishing of your own review text.
  • Using AI to check grammar or structure of your comments.

Prohibited

  • Submitting AI-generated review reports as your own expert assessment.
  • Uploading any part of a manuscript to public AI tools or services that may store or train on inputs.
  • Using AI to infer identities in blinded review.

Requirements

  • You remain fully responsible for your review’s content and tone.
  • If AI assisted in drafting or polishing your review text, disclose this in the confidential box to the editor (tool, version, purpose).

 

Editors and the journal

Permitted

  • Editorial triage (e.g., language checks, structural summaries) and workflow support using local or enterprise AI systems that do not train on confidential content.
  • Drafting decision letters for clarity (final decisions must reflect editor judgement).
  • Similarity checks, image-integrity screening, and metadata validation using approved tools.

Prohibited

  • Delegating editorial decisions to AI.
  • Uploading submissions or reviews to public AI systems that may store or train on content.

Requirements

  • The journal will keep a record of approved AI tools and their data-handling terms to ensure confidentiality and compliance with ethical standards. Approved tools may include local, institutional, or enterprise versions that provide reasonable safeguards against data retention or model training. Editors and reviewers must ensure that any AI use aligns with these standards.
  • Where AI-assisted screening is used, editors will apply human judgement before any decision.

 

Disclosure examples

Authors (place in Acknowledgements/Methods)

AI assistance was used for [language editing/outline drafting/coding assistance/figure enhancement] using [Tool, Version] on [dates]. All outputs were reviewed and verified by the authors.

Reviewers (confidential to editor)

I used [Tool, Version] to improve the grammar/clarity of parts of my review. No manuscript text or data were shared with public systems that train on inputs.

 

Data protection and tool choice

  • Use only local, enterprise, or no-train AI systems with explicit contractual guarantees that prohibit data retention or model training.
  • Never upload personal, clinical, proprietary, or embargoed material to public AI tools.
  • All AI use must comply with approved ethics protocols and data-sharing consents.

 

Non-compliance

Failure to comply may lead to manuscript rejection, corrigendum/retraction, review removal, or notification to institutions/funders, consistent with COPE guidance.

 

Definitions

  • AI: Software that generates or transforms text, code, images, audio, or data patterns.
  • Public AI tool: A service that may log or train on user inputs unless explicitly disabled by contract or setting.
  • Material use: Assistance that substantively shapes text, analyses, figures, or conclusions.