Artificial Intelligence (AI) & Generative AI Policy

ARCITECH: Journal of Computer Science and Artificial Intelligence supports the responsible, transparent, and scientifically rigorous use of artificial intelligence while protecting research integrity, confidentiality, and human accountability. This policy applies to authors, corresponding authors, peer reviewers, editors, editorial staff, and persons acting on behalf of the journal.

Generative AI refers to systems capable of producing new text, software code, images, audio, synthetic data, or other content (including large language models and AI-based image generators). AI-assisted research includes algorithms and models studied, developed, trained, or applied as part of the scientific method. Automated editorial tools include software used for similarity screening, metadata checks, image screening, or reviewer suggestions.

Human authors and journal decision-makers remain fully responsible for the accuracy, validity, originality, traceability, ethical compliance, and final scholarly content of their work. AI output is not independent scientific evidence and cannot replace critical reasoning, direct examination of primary sources, rigorous evaluation, or human judgment.

• Transparency: material use of generative AI is disclosed clearly and specifically.
• Verifiability: statements, calculations, citations, code, visualizations, and outputs are checked against reliable evidence.
• Reproducibility: computational methods and AI components relevant to research findings are described sufficiently for scientific assessment.
• Privacy and rights: personal, confidential, licensed, and sensitive data are protected.
• Accountability: authors, reviewers and editors—not AI systems—are answerable for scholarly and editorial decisions.

AI-assisted spelling correction, grammar improvement, style polishing, formatting, and translation of text originally drafted by the authors are generally permitted. Disclosure is not normally required for straightforward language correction, editing or formatting that does not introduce substantive content.

Authors must nonetheless verify the accuracy of translations and ensure that technical definitions, mathematical statements, data, citations, and scientific interpretations remain faithful to the original research. Any rewriting that changes or produces substantive scientific arguments, literature synthesis, results, or interpretation is treated as material AI assistance and must be disclosed.

Generative AI may assist with research planning, explanation, outlining, drafting, paraphrasing, literature organization, code generation or content preparation only when authors retain substantive intellectual control and critically verify the resulting material. Material generative AI assistance must be declared in the submission and in a clearly identified statement within the manuscript.

Disclosure must describe the tool or service, provider (where identifiable), model/version when relevant, purpose, affected content or workflow, and the authors’ verification and editing. Generic statements such as “AI was used” are insufficient when AI materially influenced the manuscript. Authors must not conceal extensive AI generation as routine copyediting.

AI systems, chatbots, models and automated tools must not be named as authors or co-authors. Only accountable human contributors can accept authorship responsibilities, approve the final version, respond to ethical inquiries, and manage intellectual-property and licensing obligations.

Generative AI responses must not be cited as scholarly source evidence in place of traceable primary literature or reliable data. Every cited paper, DOI, quotation, dataset, software dependency and factual statement must be independently verified. Fabricated references, invented quotations, misrepresented contributions, or untraceable research claims are unacceptable even when produced by an AI tool.

If a manuscript studies a particular model or generated response as a research object, the authors may identify the tool/model and document the output as part of the experimental materials, without presenting the system as a conventional scholarly authority.

When AI or machine learning is used to collect, preprocess, generate, classify, analyze, simulate, or interpret research data, authors must report the methodology with sufficient clarity to permit critical evaluation. Where material to the research question, this includes:

• Name, architecture, model family, provider, relevant version and access date for external models.
• Dataset origin, inclusion/exclusion criteria, sampling strategy, transformations, licenses and permissions.
• Train/validation/test partitions; controls against data leakage and inappropriate test-set reuse.
• Prompts, templates, decoding settings, seeds or run configurations where these materially affect reported findings and where lawful to share.
• Metrics, baselines, uncertainty assessment, statistical procedures, robustness tests and known limitations.
• Model checkpoints, software libraries, code versions, hardware or compute resources when needed for interpretation and reproducibility.

Reproducibility does not require public disclosure of passwords, private data, proprietary secrets, or materials whose release would create substantial security or privacy risks. Authors should explain any reasonable restrictions and provide a verifiable alternative where practicable.

Synthetic data, augmented records, AI-generated labels, simulated images, or model-generated text may be valid research materials when their provenance and intended role are clearly disclosed. Such data must not be passed off as real-world observations, independently measured results, or genuine human responses.

Authors should distinguish empirical data from synthetic or augmented data; describe how synthetic samples were generated and validated; report possible distribution shift, memorization, leakage, representativeness, and bias; and assess the consequences for generalizability.

AI-generated annotations, labels or literature classifications should be subject to appropriate quality assurance, including human review or independently validated evaluation when relevant.

AI-assisted programming is allowed, but authors remain responsible for the correctness, provenance, security, licensing, and scientific validity of all submitted code and experimental pipelines. Generated code must be tested and reviewed for hidden dependencies, insecure behavior, license conflicts, bias and erroneous outputs. Material AI assistance in developing algorithms or experimental scripts should be disclosed in the Methods or AI Use Declaration.

For cybersecurity, network, digital forensics and systems research, authors must observe applicable law, authorization requirements, responsible disclosure, and safeguards protecting systems, participants and data. AI-generated penetration-testing content or security scenarios must not be reported as real attacks or observations unless supported by verifiable evidence and appropriate authorization.

Figures, tables, diagrams and images must accurately represent the underlying data and methods. AI-assisted enhancement that changes scientific meaning, conceals relevant features, creates false patterns, or fabricates observations is prohibited.

Illustrative AI-generated graphics may be used where ethically appropriate and explicitly labeled as synthetic or AI-generated, with sufficient tool and purpose disclosure. They must not be presented as photographs, microscope images, experimental captures, screenshots or other primary evidence.

In studies where image synthesis, computer vision or deepfakes are themselves the research subject, generated outputs may be reproduced as experimental artifacts, clearly identified and ethically handled.

Authors must not input confidential, personally identifiable, unpublished third-party, contractually protected, or otherwise restricted material into AI services unless lawful authority, appropriate permissions, and adequate information-security and data-handling safeguards exist. The same principle applies to protected research datasets and proprietary source code.

Authors must respect copyright, database rights, attribution, license obligations and relevant consent requirements. Using AI to paraphrase or regenerate protected content does not eliminate the duty to acknowledge sources or obtain necessary permissions.

Reviewers must personally read, analyze and evaluate assigned manuscripts. They must not use generative AI to produce peer-review assessments, recommendations, methodological judgments, or substantive reviewer comments. Such evaluations must reflect independent expert judgment.

Reviewers must not upload submitted manuscripts, supplementary files, unpublished data, images, reviewer correspondence, or other confidential information to publicly accessible or third-party AI systems. Any proposed exception requires prior written permission from the journal and the necessary author consent and verified safeguards; permission does not authorize AI-generated scientific judgments.

Routine language polishing of a reviewer's own independently written comments may be acceptable only if no confidential content is exposed, it follows the journal's instructions, and the tool use is disclosed to the editorial office where material. Reviewers must inform the journal of any AI assistance relevant to their review and remain fully accountable.

Editors and editorial staff must not delegate scientific judgment, manuscript acceptance/rejection, reviewer evaluation, or the drafting of substantive editorial assessments to generative AI. Final decisions remain with an authorized human editor and follow ARCITECH's Peer Review Policy & Editorial Decision.

Confidential submissions, authors' personal information, reviewer identities and reports must not be supplied to external AI services unless the journal has verified confidentiality, legal compliance, appropriate permission and secure processing arrangements. Editorial workflows that use AI-assisted language editing must preserve confidentiality and disclose material use.

Editors must not use AI systems to infer sensitive personal characteristics, favor an author's identity or reputation, or generate unverified accusations of misconduct.

Journal or publisher use of automated tools for similarity screening, metadata checks, image-integrity screening, reviewer discovery, or administrative processing must be appropriately evaluated and supervised by trained personnel. Any routine use that affects editorial workflows should be transparently communicated to authors.

Software flags and predictions require proportionate human verification. Automated recommendation, integrity detection, or screening is not itself a final scientific or ethical determination. Editors must be able to explain the evidence underpinning a decision, independent of a proprietary model score.

ARCITECH will not claim that any particular AI integrity-detection software has been deployed unless its implementation is documented and operational.

AI-generated-text detection systems can yield false positives and false negatives. An AI-detection score alone is not sufficient evidence of misconduct and must not serve as the sole basis for rejection, correction or retraction.

Where concerns arise, editors should examine the manuscript and supporting evidence, verify bibliographic and data integrity, request a reasonable explanation or underlying records from the authors, and provide an opportunity to respond. Screening decisions should be documented and reviewed by a qualified human editor.

The journal’s existing similarity-screening practices remain separate from the assessment of whether AI was used and whether such use was appropriately disclosed.

Authors who have used generative AI materially during research or preparation must add a separately headed “AI Use Declaration”, preferably before the References, and disclose material use at submission. If no material generative AI was used, authors may state this explicitly for transparency; the journal does not require an AI declaration for ordinary spelling or grammar checking.

Failure to disclose material AI use, inclusion of fabricated references or results, manipulation of data or figures, or violation of manuscript confidentiality may trigger an editorial inquiry in accordance with the journal’s Publication Ethics and Peer Review policies.

Possible actions, depending on the evidence, severity, stage of publication and the authors’ response, include requesting clarification or a corrected declaration, requiring independent verification or revision, declining a manuscript, or issuing a post-publication correction, expression of concern or retraction when justified.

Allegations will be assessed by an independent, appropriately qualified editor. Authors should be informed of substantive concerns and given a reasonable opportunity to respond. Sanctions are not automatic and an unverified AI-detector prediction is insufficient grounds for a misconduct finding.

This policy complements ARCITECH's Peer Review Policy & Editorial Decision, Publication Ethics, and Author Guidelines. Where clarification is necessary, authors should contact the journal before submission.

Editorial contact: arcitech.journal@iaincurup.ac.id.

Effectiveness: this policy becomes effective prospectively when formally approved and published by the journal. Subsequent substantive updates will be announced on the journal website. The journal will periodically review the policy as AI technologies, scholarly norms and applicable standards develop.

This policy was drafted with reference to the following public guidance:

• DOAJ — Guide to Applying, Version 2.9: Artificial Intelligence (AI) and Automated Tools.
• Committee on Publication Ethics (COPE) — Authorship and AI Tools.
• International Committee of Medical Journal Editors (ICMJE) — Use of Artificial Intelligence in Publishing.
• Association for Computing Machinery (ACM) — Publications Policy on Authorship.
• Principles of Transparency and Best Practice in Scholarly Publishing.

References indicate an editorial framework, not endorsement, certification, DOAJ inclusion or formal accreditation of ARCITECH by any named organization.