1. Artificial intelligence and authorship

AI tools and other software systems cannot be authors or co-authors. Only natural persons meeting the journal’s Authorship Principles may be included in the byline.

AI cannot fulfil an author’s duties:

• accepting personal responsibility for the work’s content and integrity;
• approving submission and the final manuscript;
• disclosing its own competing interests;
• giving authorial consent to the transfer or licensing of rights;
• entering a publishing agreement as an author;
• taking responsibility for resolving questions about accuracy and provenance;
• participating as an accountable person in an investigation, correction or retraction.

A tool developer, account holder or person entering prompts does not automatically qualify for authorship either. Actual intellectual contribution and fulfilment of all criteria are assessed. Tool information belongs in the method description or AI declaration, not the byline.

2. Use of artificial intelligence by authors

Subject to this policy, authors may use AI for:

• language editing, translation and improving accessibility;
• preliminary outlines, structure and draft wording;
• searching and organising literature;
• programming assistance and data processing;
• documented analysis, classification and modelling;
• producing reproducible visualisations;
• organising research materials.

Authors independently verify:

• facts, scientific claims and the accuracy of legal terminology;
• calculations, formulas, statistical estimates and their limitations;
• the existence of sources, accuracy of quotations and whether references support claims;
• code functionality and the meaning of algorithms;
• the quality of analysis and interpretations;
• images, tables, maps and graphs;
• consistency of conclusions with the results actually obtained;
• rights to materials and compliance with licences.

AI must not be used for:

• fabricating or falsifying data, interviews, participant responses, criminal case files, judicial decisions, statistics, references or evidence;
• concealing plagiarism, reuse, duplicate publication or manipulation of screening;
• creating knowingly misleading interpretations or conclusions;
• replacing authors’ independent scientific judgement;
• presenting assumptions, generated documents or quotations as verified facts;
• distorting tables, crime maps, interview records or other results;
• improperly transferring personal, official, confidential or restricted materials;
• concealing a substantial role of AI that requires disclosure.

An AI response must not replace a primary source supporting a fact or legal proposition. Authors check the document itself and its applicable version. Where model responses are the subject of research, they may constitute research material if their provenance is clear; they must not be presented as a person’s actual testimony or confirmation of an event.

In literature reviews, AI does not replace a reproducible search and selection criteria where the review type requires them. Authors verify retrieved publications and describe substantial AI use in searching, screening or data extraction.

3. Disclosure of artificial intelligence use

Disclosure is required for generative AI used to create or substantially revise text, translate, prepare code or images, or support analysis and interpretation. Machine learning used as a research method is described regardless of whether generative AI helped write the manuscript.

Disclose:

• the tool and model name, where known;
• the version, or the date or period of use if the version is unavailable;
• the developer or provider where needed for identification;
• the purpose and specific task;
• the affected sections, data, code, figures or stages;
• the extent of AI involvement and substantive human verification;
• material tool limitations affecting assessment of the work.

Information is provided at submission in the cover letter or designated system field and in the manuscript itself. Research use is described in Methods; writing assistance is reported in a separate AI Use Declaration. Figures also carry a caption under section 4. An acknowledgement does not replace a method description. Substantial use is disclosed in the published article and updated after revisions.

Routine spelling, punctuation, grammar or reference-format checks and minor editing without generating or substantially changing content generally do not require a separate declaration. New passages, substantial restructuring and generative translation must be disclosed. The actual function, rather than the software’s name, determines the requirement; substantial use of an embedded feature also requires disclosure.

For double-anonymous review, necessary methodological details remain in the anonymised manuscript; identifying organisational details may be submitted separately where needed. Anonymisation must not conceal AI’s role, and an unknown version must not be reported as established.

Example declaration for writing assistance

In preparing the manuscript, the authors used [tool, model or version; if unavailable, date] for [specific task and affected sections]. The authors checked the output as follows: [description of checks actually performed]. They revised the output and accept responsibility for the accuracy, originality and integrity of the final version.

Example description of research use

At the [stage] stage, [tool, version or date] was used for [task] on [description of permissible data]. Methods describe the settings and procedure; output quality was assessed through [verification or validation]. Material limitations: [limitations]. Availability of code and necessary records: [location and access conditions or justified restrictions].

Where no use requiring disclosure occurred

The authors declare that they did not use AI tools in conducting the research or preparing the manuscript for purposes requiring disclosure under the journal’s policy.

Statements are adapted to actual practice. Authors must not claim checks that were not performed or use a negative declaration when substantial AI use occurred. All authors are responsible for complete disclosure; the corresponding author coordinates the final wording.

4. AI-generated images, figures and visual materials

A generated or reconstructed image must not be presented as an authentic photograph, observation record, document, physical evidence or actual measurement. In criminology, this also applies to incident scenes, participant images, document screenshots and crime maps.

The following cases are distinguished:

• Explanatory illustration: explains a concept or process rather than displaying collected data; it is checked by the author and clearly labelled as AI-generated or AI-assisted.
• Data visualisation: is derived from real or separately identified synthetic data using verifiable code; the figure must accurately reflect the data, scales, categories and uncertainty estimates.
• AI as part of the method: image processing or generation is a documented subject or stage of research; original and transformed materials are distinguished and the procedure’s scientific validity is assessed.

For permitted use, the caption identifies the tool, its role and the nature of changes; the research procedure is described in Methods. Necessary originals, code and transformation records are retained. Use must respect material rights, participant consent, confidentiality and licensing conditions.

Scientifically significant features must not be misleadingly added, removed, concealed or duplicated. An “AI-generated” label does not itself authorise use of another person’s image or an identifiable likeness. Illustrative material must not create a false impression about an actual person or event.

5. AI in programming, modelling and data analysis

Authors must not accept generated code, calculations, classifications or predictions without checking them. Methodological detail is proportionate to the tool’s role and includes, where applicable:

• the source, composition, quality and permissible use of data;
• the task, tool, model, version or period of use;
• preprocessing, features, coding rules and material settings;
• significant prompts, instructions, generation parameters and output-selection procedures;
• separation of training, validation and test data and safeguards against leakage;
• validation on independent or appropriately held-out data, suitable metrics and uncertainty;
• comparison with reference annotations, expert assessment or a justified baseline method;
• assessment of errors, systematic bias and limits on generalisation;
• availability of code, versions and sufficient records, or an explanation of restrictions.

For offence classification, judicial decision analysis or risk assessment, authors explain the target variable and annotation quality. Recorded crimes or official decisions must not be equated without justification with all acts actually committed. Where justified, differences in error rates between groups are assessed without improperly disclosing sensitive characteristics. A prediction must not be presented as a causal finding or established individual culpability.

Qualitative research requires checks of transcription and translation accuracy, preservation of participants’ meaning, coding procedures and the researcher’s role in interpretation. A model must not “reconstruct” a missing statement and present it as a participant’s words. An automated label is not validated merely because another model agrees with it.

Synthetic data, simulated scenarios and generated responses are permissible as a clearly identified research object or method. They are distinguished from empirical observations, do not increase the reported number of real participants and do not silently replace actual information.

Material input instructions, versions, settings, outputs and human decisions are retained to the extent needed for assessment. Open sharing does not require disclosing personal data, closed case files or protected prompts. If changes to a service prevent exact repetition, this is explained and the greatest practicable verifiability is provided; reproducibility must not be promised without grounds.

6. Use of artificial intelligence by reviewers

Reviewers personally perform substantive scientific assessment. Generative AI must not be tasked with writing a scientific report, assessing the validity of conclusions in place of the expert or recommending acceptance or rejection. Subsequent human editing of an AI-generated review does not make such delegation permissible.

Manuscripts and excerpts, supplements, data, code, invitations, review reports, participant information and confidential correspondence must not be uploaded to public or journal-unauthorised AI services. Fabricated criticisms, sources or evidence and misuse of confidential information for others’ interests are prohibited.

Limited language, administrative or accessibility assistance is possible only where:

• the editorial office has approved the specific tool, function and processing arrangements in advance;
• access has a permissible basis and confidentiality is protected;
• the tool does not generate substantive scientific assessment;
• the reviewer checks the final text and informs the editor of the tool, purpose and nature of assistance.

A reviewer’s own opinion that a service is safe does not replace approval. Ordinary statistical software and reproducible analytical tools may be used to check methods within the reviewer’s competence and authorised access; they do not replace expert judgement. AI does not count as one of the independent external reviewers required by the Peer Review Policy.

7. Use of artificial intelligence by editors

An authorised editor performs scientific and editorial assessment. Generative AI is not used to create substantive evaluations, analyse review reports in place of the editor, recommend acceptance or rejection, determine culpability or independently decide a complaint.

Editors must not transfer protected materials to public or unauthorised systems. Limited administrative or language assistance is permissible in a pre-approved protected environment, such as editing the wording of a decision already made by a human without changing its grounds. The tool, task and verification are recorded.

Discriminatory selection of authors or reviewers, concealed sponsor influence and bypassing recusal through an automated tool are prohibited. Neither a position nor system access authorises processing others’ data outside the assigned task. A conflicted case is referred to an independent person under the Editorial Governance and Editorial Independence Policy.

8. AI and automated tools used by the journal and publisher

The founder and publisher is the Research Institute of Criminology of the Republic of Uzbekistan. The journal, Institute and authorised staff may use assessed supporting tools for:

• identifying text similarities;
• checking bibliographic details and DOIs;
• detecting technical indicators of image alteration;
• language editing and accessibility;
• preparing and checking metadata;
• checking format and completeness;
• preliminary identification of possible reviewers;
• detecting signals requiring integrity assessment;
• technical publication preparation;
• website and editorial system administration.

These functions do not authorise generative AI to perform the substantive assessment prohibited by sections 6–7. A human verifies potential reviewers’ identity, expertise, independence and conflicts. Technical staff do not acquire scientific decision-making authority.

An automated result alone must not cause acceptance, rejection or removal from consideration, a substantive published correction, retraction, a plagiarism or other misconduct finding, or a sanction. An authorised person checks context, grounds, limitations and error risk, and the decision is documented.

AI functions actually used by the journal and their role in significant editorial processes are disclosed on the website; the description reports actual practice rather than every potential capability. Where AI materially affects published content, its use is disclosed and checked, and authors approve changes to their material. Mentioning a service, licence or integration in this policy does not establish that it is operational.

9. Confidentiality, privacy and data protection

Unpublished manuscripts, reviewer identities and reports, correspondence, personal data, restricted research materials, intellectual property and official information must be protected. Particular care is required for victims, minors, witnesses, convicted persons, informants and other dependent participants.

Before permitted processing, verify:

• authority to access and transfer data, necessary consent and ethical conditions;
• the minimum necessary information and re-identification risks;
• retention, deletion, provider staff access and third-party transfer arrangements;
• prevention of unauthorised model training or improvement using the data;
• rights to inputs and outputs;
• consistency of the specific settings and version with agreed arrangements;
• an accountable person and an incident response procedure.

Administrative permission or journal approval does not replace participant consent or other grounds where required. A paid plan, “private” label or local deployment does not itself demonstrate safety. Removing names may not prevent identification from case details. If conditions cannot be met, materials must not be transferred to the tool.

If a leak is suspected, use is paused, further access is restricted and the responsible editorial person is notified promptly. Necessary records are preserved; consequences and appropriate notifications are assessed under applicable requirements. Sensitive information must not be transmitted again merely to demonstrate the breach.

10. Human oversight and accountability

Oversight must be substantive: the responsible person understands the task, can verify the output, correct errors and explain the decision’s grounds. Simply agreeing with a system’s answer is insufficient.

Authors, reviewers, editors and staff, within their roles:

• verify factual, scientific and technical accuracy;
• assess errors, systematic bias and omissions;
• check quotations and references against authentic materials;
• verify calculations, code, visualisations and interpretations;
• protect confidentiality and third-party rights;
• document substantial AI assistance;
• accept responsibility for the final outcome.

An AI detector percentage does not prove authorship, plagiarism or misconduct. The result is assessed against specific circumstances; authors are not required to “make the score zero”. An accusation must not rest solely on style, language proficiency or a detector signal. Another model’s response also does not replace independent evidence.

A tool is not an independent author, reviewer, investigative authority or authoritative source of facts. Where verifiability is insufficient, editors may request clarification and make a reasoned quality decision without declaring misconduct unsupported by evidence.

11. Non-compliance

Failure to disclose substantial use, fabrication, concealed appropriation, manipulation of assessment and improper data transfer are handled under the Policy on Handling Allegations of Misconduct and other applicable rules. Permitted AI use is not itself grounds for a sanction.

Depending on the evidence, action may include:

• requesting explanations and permissible supporting information;
• completing a declaration or correcting the method description;
• pausing consideration for further checks;
• independent technical or ethical assessment;
• rejection or withdrawal of acceptance on sufficient grounds;
• correction of a published article;
• an expression of concern over a serious unresolved issue;
• retraction for material unreliability or another sufficient ground;
• proportionate restrictions on editorial or reviewing responsibilities;
• referral to a competent institution, ethics committee, funder or other authorised body.

Honest error, incomplete disclosure, negligence and intentional misconduct are distinguished. Authors and others concerned are normally given specific concerns and an opportunity to respond; urgent interim measures and exceptions are documented. Cases involving an editor or the Institute are handled independently, without control by an interested person.

A missing AI declaration alone does not automatically require retraction. Publication action depends on reliability, ethical acceptability and the possibility of accurate correction. An online article is considered published; substantive changes and notices retain a traceable history. Manuscript withdrawal does not automatically end a separate assessment.

12. Complaints and appeals

Grounds may include:

• an incorrect assessment of actual AI use;
• a false-positive automated signal;
• a confidentiality breach or improper material transfer;
• failure to follow the published policy;
• a decision without sufficient human checking;
• a conflict of interest or material factual or procedural error.

The Complaints and Appeals Policy applies. A submission identifies the specific issue, evidence and effect of the error; repeated disagreement without grounds is insufficient. An independent person uninvolved in the original decision conducts reconsideration.

An ordinary editorial appeal is submitted within 30 calendar days of receipt of the decision, with justified exceptions. Material new information and serious ethical risks are handled under reopening rules; concerns about published material’s reliability are not excluded by a fixed reporting limitation period.

The acknowledgement target is 10 working days; an initial response or explanation of next steps is provided where possible within 30 calendar days; ongoing cases are normally updated at least once every 30 calendar days. These are communication targets, not a guaranteed investigation deadline. Good-faith reporters are protected against retaliation; no separate fee is charged.

13. Policy administration

The Editor-in-Chief oversees implementation except in cases involving them or a conflict, when an independent authorised person acts. The Research Institute of Criminology of the Republic of Uzbekistan provides necessary organisational support and security while preserving scientific and editorial independence.

The policy takes effect after approval and website publication. Actual approval, effective and revision dates are stated in the published version; language versions are updated together. Technological and practice changes inform review. Past conduct is assessed with regard to the rules applicable at the time; new provisions are not represented as having applied historically.

Contact: current editorial contacts on the journal website; the founder and publisher’s general address for forwarding a report is info@uzkti.uz. The initial message must not contain closed case files, participant identifiers or confidential prompt logs. Describe the issue first and agree a secure channel; request independent referral where a conflict exists.

This policy operates alongside the journal’s rules on ethics, authorship, plagiarism, peer review, editorial independence, misconduct allegations and appeals. Methodological guidance includes COPE Core Practices and general ICMJE provisions on AI for authors and reviewers. They do not replace this journal’s specific rules or imply membership or indexing.