Generative AI Policy for Editors

DEPLOY recognises that generative AI tools can assist editorial work, but editors hold two responsibilities that cannot be delegated to a machine: protecting manuscript confidentiality and making editorial decisions. This policy follows COPE guidance.

1. Confidentiality Is the Hard Limit

  • Editors must not upload manuscripts, abstracts, figures, supporting data, review reports, or decision letters to public generative AI tools.
  • The prohibition covers both full and partial content, including short excerpts conveying core ideas.
  • The only exception is AI tooling officially provided by the publisher with contractual guarantees that data is neither retained nor used for model training.

2. Editorial Decisions Must Be Made by Humans

  • Desk rejection, revision, acceptance, and rejection decisions must never be determined by AI output.
  • AI must not be used to assess scientific quality, novelty, or methodological soundness.
  • AI must not select reviewers without human verification of expertise, track record, and conflicts of interest.
  • Decision letters must be composed and verified by the editor; raw AI output must never be sent as is.

3. Permitted Uses

Editors may use generative AI for tasks that involve no confidential material and replace no scholarly judgement: language editing of public-facing announcements and web pages, drafting generic correspondence templates, translating administrative material, and assisting with metadata scripts. Similarity-detection, statistical-check, and image-integrity tools are not covered by this prohibition.

4. Accountability

Editors remain fully responsible for any AI-assisted output, including factual errors, bias, and fabricated references. Where AI use affects a manuscript, it is recorded in the OJS editorial history.

5. Suspected Undisclosed AI Use

Editors must not base decisions solely on "AI detector" scores, which are unreliable and frequently misclassify writing by non-native English speakers. Suspicions are handled by examining substantive indicators (unverifiable references, inconsistent data, irreproducible methods), requesting written clarification from the corresponding author, and deciding on the basis of that response under the Allegation of Research Misconduct procedure.