AI systems do not understand scientific claims, cannot verify empirical findings, and cannot be held accountable for errors or harm. Responsibility for accuracy, originality, scientific validity, interpretation, conclusions, and regulatory compliance rests exclusively with human authors and their institutions.

Using AI to draft, analyze, or visualize does not transfer liability to the tool, its vendor, or the platform. Authors must verify every statement, citation, statistic, figure, and conclusion before submission. Editorial acceptance does not relieve authors of post-publication accountability.

Compliance obligations — including ethics approvals, consent, funding disclosure, conflict declarations, data protection, and discipline-specific regulations — apply fully to AI-assisted work. Authors who delegate tasks to AI must still ensure those tasks meet the same standards as work performed without assistance.

Disclosure is required when AI materially contributes to any stage of the work. Examples include writing assistance beyond minor grammar correction; language editing that restructures arguments; machine translation of substantive content; data analysis or statistical assistance; programming or code generation; figure preparation or image generation; literature summarization that informed framing or claims; and review assistance that shaped manuscript revision.

Disclosures should identify the tool or class of tool used, describe the task performed, and indicate the extent of human verification. Vague statements such as "AI was used" without context are insufficient when assistance was material.

Disclosure applies during submission, may be requested during peer review, and should appear on published records where platform policy and publication type require it. Reviewers and editors who use AI in workflow tasks must comply with confidentiality and role-specific policies described in later articles.

Authorship reflects intellectual contribution, accountability, and the ability to approve the final manuscript and respond to questions about the work. AI systems meet none of these criteria. Listing a model, chatbot, or software product as an author misrepresents how scholarship is produced and violates international authorship norms.

Researchers who use AI to draft text must still ensure that listed authors meet authorship criteria defined by discipline and platform policy. Gift authorship and ghost authorship remain misconduct whether or not AI was involved.

Acknowledgements may reference AI tools where appropriate, but acknowledgement does not replace disclosure of material assistance or satisfy authorship accountability requirements.

Generally acceptable uses include grammar and spelling improvement; language refinement that does not alter scientific meaning without author verification; translation with human review of technical accuracy; coding assistance for routine scripts with author validation; data organization and formatting; reference formatting and bibliography management; and visualization assistance where the underlying data and design choices remain author-controlled.

Acceptable use still requires human verification. A grammar suggestion that changes a quantitative claim, a translation that alters methodology, or code that produces incorrect output becomes a scientific integrity issue if published without correction.

The boundary between acceptable minor assistance and material assistance requiring prominent disclosure depends on how substantially AI shaped the intellectual content. When in doubt, authors should disclose.

Material assistance includes AI-generated or AI-restructured prose in introduction, methods, results, or discussion sections; automated literature reviews that informed the research question or framing; statistical or analytical outputs used without full manual replication; generated or substantially altered figures, diagrams, or images; code that performs core analysis without line-by-line author verification; and machine translation of entire sections without expert human review.

Disclosure expectations include naming the tool, describing the task, stating which sections were affected, and confirming human verification steps. Authors should explain limitations introduced by AI assistance, such as unverified citations or unvalidated code paths.

Editors and reviewers may request expanded disclosure when AI use is evident, inconsistent with claims, or relevant to evaluating reproducibility and integrity.

Unacceptable uses include fabricating data or results; inventing references or citations; generating fake peer reviews or reviewer personas; manipulating images to misrepresent findings; creating fraudulent experimental outcomes; impersonating reviewers, editors, or authors; submitting AI-generated research presented as wholly original empirical work without disclosure; and using AI to produce confidential peer-review content for external reuse.

Presenting synthetic content as human-generated primary research is misconduct. Using AI to evade plagiarism detection, duplicate-publication screening, or identity verification violates platform policy and may constitute fraud.

Unacceptable uses trigger investigation, rejection, correction, retraction, or account sanctions regardless of whether the author claims ignorance of policy.

Unpublished manuscripts are confidential. Uploading manuscript text, data, figures, or reviewer reports to public or third-party AI systems without explicit authorization violates reviewer obligations and may breach legal and ethical duties.

Limited AI assistance — such as grammar improvement of a reviewer's own comments or organizing notes — may be acceptable when no confidential content is exposed to external systems and the reviewer retains full professional judgment over recommendations.

Reviewers remain responsible for the substance of their reports. AI must not be used to generate superficial, generic, or fabricated reviews. Editors may reject reviews that appear automated, plagiarized, or inconsistent with manuscript content.

Responsible editorial uses may include drafting templated correspondence with human review; summarizing internal notes that do not contain confidential manuscript content in unauthorized external systems; and organizing submission metadata. AI must not autonomously accept, reject, or revise manuscripts.

Editors must verify AI-assisted communications for accuracy and tone before sending. Decisions on scope, integrity, ethics, and scholarly merit require human editorial judgment informed by reviewer reports and published criteria.

When editors use AI on confidential materials, they must use approved platform tools or environments that meet data protection requirements — not public chat interfaces that retain or train on uploaded content.

Models exhibit bias reflecting training data and design choices. They hallucinate plausible but false citations, statistics, and methodological details. They may reproduce copyrighted text without attribution. Authors must not cite AI-generated references without independent verification.

Reproducibility requires that AI-assisted analyses can be replicated or audited. Authors should document prompts, software versions, and validation steps where material to the findings. Human validation is mandatory for all claims presented as empirical.

Editors and reviewers should scrutinize AI-assisted work for internal consistency, impossible statistics, mismatched methods and results, and figures that appear synthetic or manipulated.

Researchers must not submit patient data, identifiable human subjects information, trade secrets, embargoed findings, peer-review content, or unpublished manuscripts to public AI services that may store, share, or train on inputs — unless contractual, legal, and institutional requirements are satisfied.

Sensitive information includes protected health information, national security–related data, indigenous knowledge shared under restricted terms, and institutional records governed by confidentiality agreements. Proprietary research shared under industry collaboration agreements requires partner consent before any AI processing.

Platform-hosted workflows should be preferred for tasks involving confidential content. When external tools are used, authors and reviewers bear responsibility for compliance with GDPR, HIPAA, institutional IRB conditions, funder data agreements, and export controls.

Transparency requires accessible disclosure statements on publication records; reader awareness of AI assistance scope; editorial records that preserve disclosure and integrity discussions; and version history that documents corrections when AI-related errors are discovered post-publication.

Review labels and trust metadata indicate what scrutiny a work received independent of AI use. AI disclosure complements — but does not replace — honest review status labeling.

When corrections arise from undisclosed or misrepresented AI use, correction notices should explain the nature of the issue without unnecessary exposure of private reviewer identity.

New modalities — agentic systems, multimodal models, autonomous research tools, and embedded laboratory AI — may blur lines between assistance and authorship. Policies must adapt without sacrificing human accountability or integrity.

Inquisra will monitor guidance from COPE, ICMJE, WAME, institutional research offices, funders, and disciplinary societies. Updates will be dated, communicated to users, and reflected in submission workflows where technically practicable.

Researchers should expect that acceptable practices may narrow or expand as evidence accumulates about AI impact on reproducibility, bias, and misconduct.

Inquisra does not prohibit legitimate AI tools. Researchers may adopt technologies that improve productivity, accessibility, and quality when use complies with this charter, institutional policy, funder terms, and applicable law.

Researchers retain intellectual property ownership subject to platform license terms. Transparent AI disclosure does not diminish authorship rights over original human intellectual contribution.

Researchers may appeal editorial decisions that they believe misapply AI policy, subject to the appeals provisions in the Editorial Ethics & Publication Governance Charter.

Reviewers may decline assignments when AI policy, confidentiality requirements, or conflicts prevent competent review. They should not be required to use AI tools that expose confidential manuscripts to unauthorized systems.

Reviewers retain credit and recognition for substantive human expert evaluation. Editorial staff should not substitute AI-generated review reports for qualified peer assessment.

Reviewers who identify suspected undisclosed AI misuse in manuscripts may report concerns through integrity channels without retaliation.

Editors interpret this charter during screening, review assignment, revision requests, and final decisions. Application must not vary arbitrarily by author reputation, institution, or geography.

Editors should request clarification when disclosures are missing, train editorial staff on evolving AI risks, and escalate suspected misconduct to integrity workflows. They must document AI-related concerns in decision history.

Editors disclose their own material AI use in communications and decisions where relevant, and recuse themselves when AI-related conflicts or commercial relationships compromise impartiality.

The platform integrates AI disclosure into submission and publication metadata; enforces authorship and integrity rules; provides education through governance documents and publication guidance; and preserves records that allow post-publication accountability.

Inquisra will not market AI features that encourage misconduct, bypass review, or misrepresent scholarly scrutiny. Innovation investments prioritize tools that help researchers verify, disclose, and reproduce work.

Trust depends on honest labels, durable records, and fair enforcement. This charter is a living commitment — not a static disclaimer — to uphold those values as technology and science co-evolve.