Purpose and Scope
This policy governs the use of generative AI, large language models, machine-learning systems, automated agents, and related AI-assisted tools in research reporting, manuscript preparation, peer review, editorial work, and publishing workflows.
Treata supports responsible innovation where tools improve scholarly work without undermining confidentiality, attribution, research integrity, intellectual property, privacy, or human accountability.
Human Accountability
Responsibility for scholarly content and consequential publication decisions remains with identifiable people. Authors are accountable for the accuracy and integrity of submitted work; reviewers are accountable for their reviews; and editors are accountable for editorial decisions.
AI output can be incomplete, fabricated, biased, misleading, improperly attributed, or unsuitable for a particular context. Human verification is therefore required wherever AI materially contributes to scholarly or editorial work.
AI Cannot Be an Author
AI systems cannot qualify as authors because they cannot take responsibility for a work, approve publication in an accountable human sense, manage conflicts of interest, respond to integrity questions as responsible persons, or enter publishing and licensing agreements.
Use of AI should therefore be described as tool use or methodological assistance where disclosure is required, not represented through authorship attribution.
Author Use and Disclosure
Authors may use AI-assisted tools where compatible with journal requirements, but remain responsible for originality, factual accuracy, citations, permissions, confidentiality, analysis, interpretation, and all submitted material.
Material use should be disclosed when required by the journal or when necessary for readers to understand how content, analysis, data, images, or methods were produced. Routine tools that do not materially affect scholarly content may be treated differently according to journal instructions.
AI in Research Methods, Data and Analysis
When AI is part of the research method rather than merely a writing aid, authors should describe the system, version or relevant configuration, purpose, inputs, validation, and limitations sufficiently for scientific evaluation where applicable.
AI-generated or transformed data must not be represented as observed empirical data. Synthetic data, model outputs, automated classifications, and algorithmic transformations should be identified accurately.
AI-Generated Images and Multimedia
Generative or transformative AI used for scientific images, figures, audio, video, or other research material must comply with image-integrity, consent, privacy, copyright, and evidentiary requirements. Generated illustrative content should not be presented as authentic clinical, experimental, or observational evidence.
Journal-specific restrictions may be stricter for clinical images, participant material, graphical abstracts, cover art, or other categories.
Reviewers and Confidentiality
Reviewers must protect manuscript confidentiality. They should not upload unpublished manuscripts, figures, data, author information, or review material to external AI systems when doing so would expose confidential content or conflict with journal, privacy, contractual, or intellectual-property obligations.
Where a journal permits limited AI assistance, reviewers remain personally responsible for the review and should comply with any disclosure requirements. AI should not replace expert assessment.
Editors and Editorial Tools
Editors and publishing staff may use approved tools for functions such as screening, language support, metadata checking, reviewer matching, or workflow assistance when appropriate safeguards exist. Confidential manuscripts should not be exposed to unapproved systems.
Automated scores, classifications, or flags must not be treated as autonomous grounds for acceptance, rejection, misconduct findings, or other consequential editorial action.
Citation, Attribution and Intellectual Property
Authors should verify citations suggested or generated by AI rather than assuming that references exist or support the stated claim. Fabricated citations, unattributed reproduction, or infringement remains the responsibility of the human submitting the content.
Use of AI does not remove obligations relating to copyright, licensing, third-party material, confidentiality, or appropriate scholarly attribution.
Privacy, Sensitive Information and Security
Personally identifiable, confidential, proprietary, embargoed, or security-sensitive material should not be supplied to AI services without an appropriate lawful basis, authorization, and safeguards. Authors, reviewers, and editors should consider whether service providers retain inputs or use them for model improvement.
Where institutional or journal-approved systems provide stronger protections, users should still comply with the purpose and limits for which those systems are authorized.
Suspected Misuse and Editorial Assessment
AI-detection tools and stylistic signals can produce false positives and should not be treated as proof that prohibited AI use occurred. Editors should evaluate the underlying evidence, request clarification where appropriate, and distinguish permitted assistance from deceptive or integrity-compromising use.
Fabricated content, manipulated data, false citations, undisclosed synthetic evidence, or other serious problems may be assessed under the relevant Research Misconduct, Data, Image, or Publication Ethics policies regardless of whether AI was involved.
Journal Configuration and Stronger Requirements
The core principles of human accountability, confidentiality, integrity, and truthful representation are publisher mandatory. Journals may impose stronger or more specific disclosure and use requirements according to discipline, article type, peer-review model, participant sensitivity, or technological risk.
Journal instructions should distinguish requirements that apply to authors, reviewers, and editors rather than using a single ambiguous AI statement for all roles.
Review Cycle and Version Control
Because AI capabilities, risks, standards, and regulation change rapidly, Treata should review this policy regularly and at least annually as a governance practice. Material revisions should receive a new version or effective date.
Earlier policy versions should remain identifiable so that requirements applicable to historical submissions can be determined.
Version Control
Version 1.0. This page is the canonical publisher-level text for the current version. Material revisions should be versioned and dated, while superseded versions remain identifiable so that historical applicability can be determined.