Food Systems, Safety & Sustainability (FSSS) recognises that artificial-intelligence and automated tools may support legitimate research, analysis, writing, translation, visualisation, and publishing activities. Their use must nevertheless preserve human accountability, research integrity, confidentiality, transparency, reproducibility, intellectual-property rights, and the reliability of the scholarly record.

This policy applies to generative artificial intelligence, large language models, chatbots, image-generation systems, code-generation tools, automated translation systems, machine-learning applications, algorithmic research tools, and other systems that generate, transform, classify, analyse, recommend, or evaluate scholarly content.

The use of an AI tool does not reduce the responsibilities of authors, reviewers, editors, or publisher personnel. Human participants remain responsible for every decision, statement, analysis, citation, image, dataset, review report, and editorial outcome produced with AI assistance.

1. Scope

This policy applies to:

• all authors and contributors;
• all manuscript and article types;
• research planning and study design;
• data collection, generation, processing, and analysis;
• literature searching and evidence synthesis;
• manuscript drafting, editing, translation, and formatting;
• preparation of tables, figures, images, code, and supplementary files;
• peer review and editorial decision-making;
• production, metadata preparation, and publication;
• post-publication assessment and research-integrity investigations; and
• automated tools used by the journal or publisher.

2. Definitions

Artificial Intelligence

Artificial intelligence refers broadly to computational systems capable of performing tasks involving prediction, classification, generation, optimisation, interpretation, recommendation, pattern recognition, or other functions commonly associated with human cognitive activity.

Generative Artificial Intelligence

Generative artificial intelligence refers to systems that generate new text, images, audio, video, code, data, summaries, translations, models, or other content in response to instructions or input.

AI-Assisted Tool

An AI-assisted tool is a system that uses artificial-intelligence techniques to support, modify, analyse, evaluate, or automate part of a research or publication process.

Substantive Use

Substantive use occurs where an AI system materially affects the intellectual, methodological, analytical, evidentiary, interpretive, or communicative content of a manuscript.

Routine Technical Assistance

Routine technical assistance includes basic spelling correction, grammar correction, punctuation correction, reference-format conversion, file-format conversion, or mechanical layout adjustment that does not generate or alter substantive scholarly content.

3. Fundamental Principles

Use of AI in research and scholarly publishing must comply with the following principles:

• Human accountability: identifiable human authors and editors must remain responsible for all work and decisions.
• Transparency: substantive AI use must be described accurately.
• Verification: AI-generated or AI-assisted output must be reviewed and verified by qualified humans.
• Originality: AI must not be used to conceal plagiarism, duplicate publication, or misappropriation.
• Accuracy: fabricated, unverifiable, biased, or misleading output must not be included.
• Confidentiality: protected or unpublished material must not be supplied to unauthorised systems.
• Reproducibility: methodological uses of AI must be documented sufficiently for evaluation.
• Fairness: AI must not introduce or conceal unjustified bias or discrimination.
• Legality: AI use must comply with copyright, privacy, data-protection, contractual, and regulatory obligations.
• Human oversight: automated output must not replace responsible professional judgement.

4. AI Systems Cannot Be Authors

Artificial-intelligence systems, chatbots, large language models, image generators, code generators, and other automated tools cannot be listed as authors.

Such systems cannot:

• approve the submitted or published manuscript;
• accept responsibility for accuracy or integrity;
• respond independently to editorial or ethical concerns;
• disclose competing interests;
• enter into a publication agreement;
• hold or transfer copyright;
• consent to publication;
• take legal responsibility for defamatory or infringing material; or
• meet the journal’s authorship criteria.

AI tools should not be included in the author byline, acknowledgements as if they were human contributors, or CRediT contributor statements.

Where AI use requires disclosure, it should be described in the Methods section, an AI-use declaration, an acknowledgement, or another appropriate section specified by this policy.

5. Human Author Responsibility

Authors remain fully responsible for all material produced or modified using AI.

This responsibility includes:

• checking factual accuracy;
• checking methodological validity;
• verifying calculations and statistical results;
• verifying quotations and citations;
• confirming that cited sources exist and support the relevant claims;
• detecting fabricated or misleading output;
• identifying bias and inappropriate generalisation;
• protecting confidential or personal information;
• ensuring originality and proper attribution;
• ensuring compliance with copyright and licensing requirements;
• preserving the integrity of data, images, code, and conclusions; and
• correcting significant errors discovered before or after publication.

Authors cannot transfer responsibility to an AI provider, software developer, language service, research assistant, or third-party manuscript agency.

6. Uses That Normally Require Disclosure

Authors must disclose AI use where a system was used substantively for:

• generating or substantially rewriting manuscript text;
• preparing an abstract, conclusion, discussion, or literature review;
• summarising scholarly literature;
• identifying, selecting, or screening references;
• translation of substantive scholarly content;
• developing research questions or hypotheses;
• designing a study, survey, experiment, model, or analytical strategy;
• generating survey questions, interview guides, or research instruments;
• extracting, coding, or classifying data;
• qualitative coding or thematic analysis;
• statistical, computational, economic, or environmental analysis;
• generating or debugging code used in the research;
• constructing mathematical or simulation models;
• creating synthetic data;
• generating, modifying, restoring, or enhancing images;
• creating figures, diagrams, graphical abstracts, or visualisations;
• interpreting findings;
• developing policy or practical recommendations;
• checking research integrity or detecting anomalies;
• preparing responses to reviewers; or
• performing another activity that materially influenced the manuscript.

7. Uses That Do Not Normally Require Disclosure

Disclosure is not normally required for routine technical functions that do not generate or alter substantive scholarly content, such as:

• basic spelling correction;
• basic grammar and punctuation correction;
• mechanical formatting of headings, margins, or references;
• conversion between common file formats;
• ordinary reference-manager functions;
• standard word-processing functions;
• deterministic calculations performed by conventional software;
• routine statistical analysis using established software where the software is reported under normal methodological standards; and
• accessibility functions that do not generate substantive content.

Disclosure is required where a tool presented as a grammar, translation, reference, or formatting assistant also generated, expanded, summarised, interpreted, or materially rewrote scholarly content.

Authors who are uncertain whether a particular use was substantive should disclose it.

8. Required AI-Use Declaration

Where disclosure is required, the declaration should identify:

• the tool or system name;
• the provider or developer;
• the model and version, where available;
• the date or period of use;
• the purpose for which the tool was used;
• the manuscript sections or research stages affected;
• whether prompts, outputs, code, or logs have been preserved;
• the procedures used to review and verify the output;
• whether confidential, personal, proprietary, or unpublished information was supplied; and
• the human authors who supervised and accepted responsibility for the use.

9. General Disclosure Template

Artificial Intelligence Disclosure: During preparation of this manuscript, the authors used [tool name, provider, model or version] for [specific purpose]. The tool affected [identify sections, materials, or stages of work]. All generated or modified content was reviewed, verified, and revised by the authors. The authors accept full responsibility for the accuracy, originality, integrity, and final content of the manuscript.

10. Language-Assistance Disclosure Template

Artificial Intelligence Disclosure: The authors used [tool name and version] to assist with English-language editing and translation. The tool was not used to generate research data, references, analyses, interpretations, or conclusions. All revisions were reviewed and approved by the authors.

11. Research-Method Disclosure Template

Artificial Intelligence Disclosure: [Tool and model] was used as part of the research methodology for [classification, prediction, data extraction, coding, modelling, or other purpose]. The model configuration, input data, validation procedures, evaluation metrics, and human-oversight procedures are described in the Methods section. The authors independently verified the resulting output and accept responsibility for the analysis and conclusions.

12. No Substantive AI Use

Where the journal or editor requests a declaration and no substantive AI use occurred, authors may state:

No generative artificial-intelligence or AI-assisted system was used to generate the scholarly content, research data, analysis, interpretation, or conclusions of this manuscript.

13. Placement of the Disclosure

AI use should be disclosed in the location most relevant to the activity:

• methodological use should be described in the Methods section;
• writing, translation, or editing assistance should be stated in the AI-use declaration or Acknowledgements section;
• AI-generated or AI-modified figures should be identified in the figure caption;
• AI-generated code should be described in the Methods or Software section;
• synthetic data should be identified in the Methods and Data Availability Statement;
• use affecting the entire manuscript should be stated on the title page and in the published declaration; and
• additional technical documentation may be supplied as supplementary material or deposited in a repository.

14. AI as a Research Method or Object of Study

This policy does not prohibit legitimate research involving artificial intelligence, machine learning, natural-language processing, computer vision, predictive models, or automated decision systems.

Where AI is the research method, intervention, system under evaluation, or object of study, authors must report sufficient information for critical assessment.

Relevant information may include:

• system and model name;
• developer and provider;
• model version and access date;
• training or fine-tuning information where available;
• input and output specifications;
• prompts, system instructions, and prompt-development procedures;
• temperature, seed, sampling, or other relevant parameters;
• data provenance and eligibility criteria;
• preprocessing and feature engineering;
• training, validation, and test-set separation;
• procedures used to prevent data leakage;
• performance metrics and uncertainty;
• calibration and external validation;
• subgroup, bias, equity, and fairness assessment;
• human oversight and intervention;
• known limitations and failure modes;
• software, code, and computing environment;
• data and model availability; and
• changes in system behaviour during the research period.

Authors should follow an applicable reporting guideline, including TRIPOD+AI or another recognised standard where appropriate.

15. Reproducibility and Documentation

Authors using AI as part of the research process should preserve sufficient documentation to permit editorial evaluation and, where possible, reproduction.

Documentation may include:

• complete prompts and system instructions;
• prompt versions and development history;
• raw and edited outputs;
• model and software versions;
• configuration parameters;
• application programming interface settings;
• code and computational notebooks;
• random seeds;
• training, validation, and test datasets;
• human review and correction records;
• excluded or failed outputs;
• quality-control procedures; and
• dates on which the system was accessed.

The journal may request relevant documentation during peer review or an integrity assessment.

16. Dynamic and Proprietary Systems

Authors should recognise that proprietary or continuously updated AI systems may produce different output over time.

Where complete reproducibility is not possible, authors should report:

• the system version where available;
• the precise date of access;
• the full prompts and relevant parameters;
• the number of runs or repetitions;
• procedures used to select among outputs;
• the preserved original output; and
• limitations created by the proprietary or changing nature of the system.

17. AI-Generated Data and Synthetic Data

AI-generated or synthetic data must not be represented as naturally observed, experimentally measured, surveyed, interviewed, or otherwise empirical data.

Synthetic data may be used where methodologically justified, provided that authors:

• identify the data clearly as synthetic;
• explain why synthetic data were used;
• describe the generation method;
• identify the source data or assumptions;
• describe validation procedures;
• assess potential bias and distortion;
• distinguish synthetic from empirical observations;
• avoid unsupported claims of real-world validity;
• preserve relevant code, prompts, and parameters; and
• state appropriate limitations.

Fabricating research observations with AI and presenting them as genuine constitutes serious research misconduct.

18. Data Analysis and Interpretation

AI may be used to support data analysis where its role is methodologically justified and transparently reported.

Authors must not rely on an AI system to:

• select a preferred analysis solely because it produces a desired result;
• exclude observations without a justified and documented reason;
• generate unverifiable statistical output;
• conceal model instability or poor calibration;
• misrepresent exploratory analysis as prespecified analysis;
• replace subject-matter expertise;
• draw causal conclusions unsupported by the research design;
• create false certainty;
• selectively omit contradictory findings; or
• make final interpretations without qualified human review.

19. Qualitative Research

Use of AI for transcription, translation, coding, thematic analysis, sentiment analysis, or interpretation of qualitative material must be disclosed.

Authors should report:

• the material supplied to the system;
• whether participants consented to the relevant processing;
• privacy and confidentiality safeguards;
• the coding or analytical function performed;
• human review of generated themes or categories;
• procedures for assessing cultural and linguistic accuracy;
• handling of contradictory or minority perspectives;
• potential bias introduced by the tool; and
• whether identifiable information was removed before processing.

AI-generated themes or interpretations must not replace reflexive human qualitative analysis without methodological justification.

20. Literature Searching and Evidence Synthesis

Authors may use AI-assisted tools to support literature searching, screening, extraction, or synthesis, but must verify the completeness and accuracy of the process.

Authors must not:

• rely on unverified AI-generated references;
• cite articles that they have not confirmed exist;
• assume that an AI-generated literature summary is complete or unbiased;
• omit databases, search terms, eligibility criteria, or screening procedures;
• replace a reproducible systematic search with an undocumented chatbot query;
• use AI-generated descriptions of articles without consulting the original sources; or
• conceal the role of an automated screening or extraction tool.

Systematic and scoping reviews must still comply with PRISMA and any applicable extensions.

21. Citations and References

Authors are responsible for verifying every reference against the original source.

Generative-AI output must not be treated as a primary scholarly source or evidence supporting a factual, scientific, legal, policy, or methodological claim.

Authors must not:

• cite fabricated references generated by AI;
• cite an AI answer instead of the original source;
• quote AI-generated text as authoritative evidence;
• rely on an AI-generated summary without consulting the underlying publication;
• attribute unsupported statements to a real source; or
• use AI to conceal plagiarism or inappropriate paraphrasing.

Where an AI system is itself the object of study, a methodological resource, or software used in the research, authors may identify and reference the tool, model, documentation, version, or archived release as appropriate. Such a reference documents the research resource; it does not make the system an authoritative source for the manuscript’s scholarly claims.

22. Writing and Text Generation

AI-assisted drafting or rewriting is permitted only where:

• the use is disclosed when substantive;
• the authors verify all factual statements;
• the authors preserve their own scholarly judgement and voice;
• all sources are checked against the original publications;
• no confidential or restricted information is supplied improperly;
• no plagiarised or closely imitative text is included;
• the output does not misrepresent the authors’ expertise or research activity; and
• the authors accept responsibility for the final text.

Submitting a manuscript generated substantially by AI without meaningful human intellectual involvement, verification, and accountability is inconsistent with this policy.

23. Translation

Substantive machine translation or AI-assisted translation must be reviewed by an author or qualified individual competent in both the source and target languages.

Authors must verify:

• technical terminology;
• quantitative values;
• legal and regulatory language;
• participant quotations;
• culturally specific concepts;
• negation and uncertainty;
• references and proper names; and
• the equivalence of the translated conclusions.

AI translation must not be used to disguise duplicate publication or undisclosed translation of previously published work.

24. Code Generation

AI-generated or AI-assisted code used in research must be reviewed, tested, documented, and validated by the authors.

Authors should report:

• the AI tool used;
• the purpose of the generated code;
• the programming language and software environment;
• testing and validation procedures;
• known limitations;
• human modifications;
• software dependencies;
• availability of the final code; and
• licensing or intellectual-property restrictions.

Authors must not include code they cannot understand, validate, maintain, or accept responsibility for.

25. Figures, Images, and Visual Materials

AI-generated or AI-modified visual material must not misrepresent research evidence.

Authors must not use AI to:

• create experimental images that did not originate from the reported research;
• add, remove, duplicate, move, or conceal scientific features;
• generate missing portions of evidentiary images without disclosure;
• replace original sample, field, laboratory, microscopic, or participant images;
• alter results to strengthen a desired interpretation;
• create deceptive maps, graphs, or data visualisations;
• simulate participant photographs without clear identification;
• remove watermarks or ownership information unlawfully; or
• present an illustrative AI-generated image as empirical evidence.

AI-generated illustrative material may be considered where:

• it is not presented as research evidence;
• its use is relevant and necessary;
• it is labelled clearly as AI-generated;
• the tool and purpose are disclosed;
• copyright and personality rights are respected;
• the material is not misleading, discriminatory, or deceptive; and
• the responsible editor approves its inclusion.

26. Image Enhancement and Restoration

AI-assisted denoising, sharpening, restoration, segmentation, or enhancement of research images is permitted only where scientifically justified and transparently reported.

Authors must:

• retain the original unprocessed files;
• describe the software, model, and processing method;
• apply processing consistently where appropriate;
• verify that no scientific feature was fabricated or removed;
• provide original files when requested;
• distinguish measured information from predicted or reconstructed information; and
• explain how the processing affected interpretation.

27. Confidentiality, Privacy, and Data Protection

Authors must not upload confidential, identifiable, proprietary, embargoed, unpublished, or otherwise restricted information to an external AI system without appropriate authorisation and safeguards.

Restricted material may include:

• identifiable participant information;
• health, genetic, biometric, or behavioural data;
• confidential interview transcripts;
• unpublished datasets;
• commercially sensitive information;
• confidential peer-review material;
• protected Indigenous or community knowledge;
• food-safety or infrastructure vulnerability information;
• trade secrets;
• copyrighted manuscripts or books;
• third-party data subject to contractual restrictions; and
• information subject to export-control, security, or regulatory requirements.

Authors must consider whether the AI provider stores inputs, uses them for model training, transfers them across jurisdictions, or permits deletion and access control.

28. Human-Participant Consent

Where participant information will be processed using an AI system, authors must ensure that the use is consistent with:

• the participant information and consent process;
• the approved research protocol;
• the ethics committee’s requirements;
• applicable privacy and data-protection law;
• data-sharing agreements; and
• reasonable participant expectations.

De-identification does not automatically eliminate all privacy risks. Authors must consider the possibility of re-identification and unintended disclosure.

29. Copyright and Intellectual Property

Authors are responsible for ensuring that AI-assisted material does not infringe copyright, database rights, trademarks, patents, licences, contractual restrictions, confidentiality obligations, or other legal rights.

Authors must not assume that:

• AI-generated output is automatically original;
• AI-generated content is automatically free of copyright restrictions;
• a tool provider grants all rights required for publication;
• the absence of a visible source eliminates attribution obligations; or
• AI-generated resemblance to protected content is legally harmless.

The journal may request evidence concerning provenance, licensing, permission, or creation of AI-assisted material.

30. Bias, Fairness, and Discrimination

Authors must evaluate whether AI systems introduce or amplify bias relating to:

• sex or gender;
• race, ethnicity, nationality, or language;
• geography and income classification;
• disability or health status;
• age;
• institutional prestige;
• occupation or socioeconomic position;
• farming system or production scale;
• Indigenous, traditional, or community knowledge;
• food cultures and dietary practices;
• underrepresented populations; and
• other characteristics relevant to the research.

Where relevant, authors should report subgroup performance, fairness measures, error distributions, model limitations, and steps taken to reduce inequitable outcomes.

31. Dual-Use, Food-Safety, and Public-Safety Risks

AI-assisted research may create risks where it facilitates harmful applications.

Authors must disclose and assess foreseeable risks involving:

• intentional food or water contamination;
• development or misuse of biological agents;
• evasion of food-safety detection systems;
• identification of vulnerabilities in supply chains or infrastructure;
• unsafe agricultural or laboratory practices;
• harmful chemical or biological optimisation;
• circumvention of regulatory controls;
• automated misinformation affecting food safety or public health; and
• other substantial risks to people, animals, communities, or the environment.

The journal may request specialist review, risk mitigation, restricted methodological detail, modification, or rejection where risks cannot be managed responsibly.

32. Prohibited Uses by Authors

Authors must not use AI to:

• fabricate data, participants, experiments, interviews, samples, observations, or results;
• generate false ethics approvals, registrations, grant numbers, or permissions;
• fabricate or alter references;
• produce false quotations;
• conceal plagiarism or duplicate publication;
• create deceptive images or evidence;
• manipulate statistical significance or model performance;
• misrepresent synthetic data as empirical data;
• impersonate an author, reviewer, editor, participant, or institution;
• generate fraudulent reviewer identities;
• manipulate peer review;
• produce defamatory or unlawfully discriminatory content;
• circumvent ethical, legal, privacy, or safety requirements;
• produce a manuscript for sale through a paper mill or authorship marketplace;
• conceal the actual authors or contributors; or
• misrepresent the extent of human intellectual involvement.

33. Use of AI by Peer Reviewers

Submitted manuscripts and associated files are confidential and privileged scholarly communications.

Reviewers must not upload or transmit a manuscript, abstract, table, figure, dataset, supplementary file, review form, or substantial extract to an external AI system unless the journal has provided explicit written authorisation and suitable confidentiality safeguards are in place.

Reviewers must not use generative AI to:

• write or substantially generate a peer-review report;
• summarise the manuscript as a substitute for reading it;
• evaluate novelty or significance without independent assessment;
• generate methodological criticism that the reviewer cannot verify;
• recommend acceptance or rejection;
• identify weaknesses through an unauthorised external system;
• draft confidential comments to the editor;
• suggest citations without verifying their relevance and existence; or
• replace the reviewer’s own expertise and judgement.

Basic local spelling or grammar correction of a reviewer’s own report may be used where it does not transmit confidential manuscript information or generate the substantive assessment.

Any authorised AI use must be disclosed to the editor. The reviewer remains fully responsible for the report.

34. Reviewer Declaration

Reviewers may be required to confirm:

I have not uploaded the manuscript or associated confidential material to an unauthorised artificial-intelligence system. I have not used generative AI to replace my independent assessment or generate the substantive content of this review.

35. Use of AI by Editors

Editors must preserve confidentiality, independence, and human responsibility throughout editorial assessment.

Editors must not use external generative-AI systems to:

• create substantive editorial assessments;
• make acceptance or rejection decisions;
• replace review of the manuscript;
• summarise confidential manuscripts without authorisation;
• generate decision letters containing criticism that the editor cannot verify;
• evaluate ethical concerns without human investigation;
• rank manuscripts through undisclosed automated scoring;
• select reviewers without human verification; or
• upload confidential author or reviewer information to unauthorised systems.

Editorial decisions must be made by authorised human editors.

36. Approved Editorial and Administrative Tools

The journal or publisher may use tested automated tools to support limited technical or administrative functions, including:

• similarity screening;
• reference and DOI checking;
• image-duplication or manipulation screening;
• metadata validation;
• language and formatting checks;
• spam, security, and fraud detection;
• reviewer discovery or expertise matching;
• file conversion and accessibility processing;
• production quality control; and
• other documented support functions.

Where such tools are used:

• their purpose must be legitimate;
• the tools should be tested and evaluated appropriately;
• confidentiality and data protection must be considered;
• relevant output must be reviewed by authorised personnel;
• false positives and false negatives must be considered;
• the tool must not make final editorial decisions;
• reviewer suggestions must be verified independently; and
• authors should be informed where automated screening materially affects editorial assessment.

37. Automated Detection Is Not Proof of Misconduct

AI-detection, similarity-detection, image-screening, or anomaly-detection output may identify material requiring further examination, but it does not by itself establish misconduct.

The journal will not normally reject, accuse, correct, or retract solely on the basis of an automated score or classification.

Where a tool identifies a concern, the journal may:

• conduct a human review;
• examine the relevant manuscript content;
• compare the material with available sources;
• request clarification from the authors;
• request original files, data, prompts, code, or images;
• obtain independent expert advice;
• allow the authors a reasonable opportunity to respond; and
• determine an editorial response from the complete evidence.

38. Undisclosed or Inappropriate AI Use

Failure to disclose required AI use does not automatically prove fabrication or other serious misconduct. The journal will consider:

• the nature and extent of the use;
• whether disclosure was clearly required;
• whether the output was reviewed and verified;
• whether the use affected data, analysis, evidence, or conclusions;
• whether confidential or protected material was exposed;
• whether the manuscript contains false or fabricated content;
• whether the omission appears accidental or deliberate;
• whether authors respond honestly and completely; and
• whether the reliability of the work remains intact.

39. Warning Signs

Potential warning signs may include:

• references that do not exist;
• citations that do not support the stated claim;
• confident statements unsupported by evidence;
• inconsistent terminology or unexplained changes in style;
• generic or non-specific methodological descriptions;
• invented quotations or statistics;
• impossible or internally inconsistent data;
• formulaic manuscripts reproduced across unrelated topics;
• images containing generated or repeated features;
• authors unable to explain the methods or analysis;
• unexplained code or model output;
• undisclosed synthetic data;
• confidential material processed through an unauthorised service; and
• peer-review reports containing unverifiable or fabricated claims.

A warning sign is not proof of misconduct and must be assessed in context.

40. Editorial Response before Publication

Where inappropriate, undisclosed, or unreliable AI use is identified before publication, the journal may:

• request clarification;
• require a complete AI-use declaration;
• request revision or removal of affected material;
• request original data, images, code, prompts, or outputs;
• require additional methodological reporting;
• repeat technical or peer review;
• replace an editor or reviewer who breached the policy;
• pause editorial processing;
• reject the manuscript;
• refer serious concerns to an institution or other authority; or
• take another proportionate action under the Publication Ethics Policy.

41. Post-Publication Action

Where an AI-related concern is identified after publication, the journal may:

• request an explanation and supporting records;
• publish or update an AI-use declaration;
• publish a correction or clarification;
• issue an expression of concern;
• conduct or request an integrity investigation;
• notify an institution, funder, ethics committee, repository, or other journal;
• retract the article where the findings or integrity of the work are unreliable;
• remove specific material in exceptional legal or safety circumstances; or
• take another action necessary to maintain the scholarly record.

Retraction is not normally required merely because an AI-use statement was omitted where the work remains reliable and the omission can be corrected transparently.

42. Serious AI-Related Misconduct

The following may constitute serious research or publication misconduct:

• AI-assisted fabrication or falsification of data;
• creation of fictitious participants, interviews, experiments, or observations;
• fabrication of references, ethics approvals, registrations, or grants;
• deceptive generation or manipulation of research images;
• deliberate concealment of substantial AI generation;
• AI-assisted plagiarism or copyright infringement;
• misrepresentation of synthetic data as empirical data;
• uploading protected participant data without authority;
• fabrication of reviewer identities or reports;
• AI-assisted peer-review manipulation;
• submission of mass-produced manuscripts without genuine human scholarship; and
• deliberate misrepresentation of human authorship or responsibility.

43. Third-Party Services

Authors remain responsible for AI use by manuscript agencies, language services, statistical consultants, research organisations, commercial laboratories, and other third parties acting on their behalf.

Third parties must not:

• generate manuscripts without genuine author oversight;
• fabricate data or references;
• upload confidential material without permission;
• conceal AI use from the authors or journal;
• sell AI-generated authorship positions;
• manipulate peer review;
• control author accounts without authorisation; or
• guarantee acceptance or publication.

Authors should disclose substantial professional writing, translation, statistical, coding, or AI-related assistance and its funding source where applicable.

44. Corrections during Proof Review

Authors must not use AI during proof correction to introduce substantial new text, data, analysis, citations, or conclusions without editorial permission.

Any new substantive AI use after acceptance must be disclosed to the production editor. Changes must remain subject to author verification and editorial approval.

45. Appeals and Complaints

Authors, reviewers, or editors may request reconsideration where they believe that:

• an automated tool produced an incorrect result;
• AI use was classified inaccurately;
• material evidence was overlooked;
• the policy was applied inconsistently;
• the assessment involved an undisclosed conflict of interest;
• confidentiality was breached; or
• the editorial action was disproportionate.

The request should identify the manuscript or article, explain the disputed decision, and provide relevant evidence.

Further procedures are described in the journal’s Complaints and Appeals Policy.

46. Record Retention

The journal may retain relevant records concerning AI use, including:

• author declarations;
• prompts and outputs supplied during review;
• model and software information;
• automated screening results;
• human verification records;
• reviewer or editor disclosures;
• confidentiality and security assessments;
• correspondence concerning suspected misuse;
• supporting data, code, or images; and
• investigation and editorial-decision records.

Records may be retained for legitimate editorial, auditing, security, legal, preservation, and research-integrity purposes, with access limited to authorised personnel.

47. Policy Changes

Artificial-intelligence technologies and scholarly-publishing standards continue to develop. FSSS may revise this policy in response to:

• changes in technology;
• new research-integrity risks;
• changes in legal or regulatory requirements;
• updated editorial standards;
• changes in data-protection or copyright practice;
• new reporting guidelines; and
• operational experience gained by the journal.

Authors, reviewers, and editors should consult the current published version of the policy when performing the relevant activity.

48. Standards Informing This Policy

This policy is informed by internationally recognised guidance concerning responsible use of artificial intelligence in scholarly publishing, including:

• ICMJE: Use of Artificial Intelligence in Publishing;
• COPE: Authorship and AI Tools;
• WAME: Chatbots, Generative AI, and Scholarly Manuscripts; and
• DOAJ Guide to Applying: Artificial Intelligence and Automated Tools.

Reference to these organisations means that their guidance informs the journal’s policy. It does not imply membership, certification, endorsement, or formal affiliation unless separately stated and independently verifiable.

49. Final Author Checklist

Before submission, authors should confirm that:

• no AI system has been listed as an author;
• all substantive AI use has been disclosed;
• the tool, model, version, purpose, and affected stages have been identified;
• all AI-generated content has been reviewed and verified;
• all references have been checked against original sources;
• AI output has not been cited as primary scholarly evidence;
• synthetic data are identified clearly;
• AI-generated or AI-modified images are disclosed appropriately;
• original data and images have been retained;
• AI-generated code has been tested and documented;
• confidential or personal data were not supplied without authority;
• copyright, privacy, and licensing requirements have been satisfied;
• bias, fairness, and safety implications have been considered;
• prompts, outputs, parameters, or logs have been preserved where methodologically relevant;
• the Methods section permits appropriate evaluation and reproduction;
• the conclusions reflect human scholarly judgement;
• all authors accept responsibility for the final work; and
• the manuscript complies with the Publication Ethics, Authorship and Contributorship, Data Availability, Peer Review, and Citation and Reference Style policies.

50. Contact

Questions concerning permitted AI use, disclosure requirements, confidential data, AI-generated images, research methods, reviewer restrictions, editorial tools, or suspected AI-related misconduct should be directed to the FSSS Editorial Office through the official contact information provided on the journal website.

Correspondence concerning an existing manuscript or article should include:

• the manuscript title;
• the submission identification number or DOI;
• the AI tool, provider, and model or version;
• the purpose and extent of use;
• the sections or research stages affected;
• the procedures used for human review and verification; and
• any relevant supporting documentation.