03 · What You Need to Know
Coordinating AI Governance Across Countries and Research Institutions
Why International AI Governance Is More Than Tool Selection
International research collaborations may involve different institutional policies, legal frameworks, disciplinary conventions, and research ethics procedures.
AI adds another layer because information may be processed through external providers, cloud infrastructure, connected applications, or locally deployed systems.
The same tool may operate under different contractual arrangements for different collaborators.
For example, a university-managed AI account may have processing protections that do not apply to another researcher's personal account.
Research teams should therefore evaluate AI use as part of the project's overall governance rather than simply agree on a preferred chatbot.
Start With a Map of the Research Collaboration
Before establishing AI permissions, the team should identify the participating institutions, their roles, and the information they contribute.
Relevant questions include where data are collected, who is authorized to access them, which institutions govern their use, and whether information will be transferred across borders.
The team should also identify the research activities in which AI may be involved.
For example, public literature exploration presents different governance considerations from processing identifiable participant interviews.
A clear project map helps researchers determine which activities require additional authorization.
Identify the Applicable Institutional Requirements
Each participating institution may establish rules concerning AI access, information security, research ethics, and scholarly disclosure.
These requirements may differ in scope.
One university may restrict particular external services, while another may permit them only for nonconfidential material.
Researchers should determine which rules govern their personnel, systems, and contributed research information.
The question of conflicting institutional AI policies in research should be resolved through actual authority and applicable agreements rather than assumptions that the lead institution always prevails.
Which Legal and Governance Issues Should Be Considered?
| Governance issue |
Question for the research team |
Possible implication |
| Personal data protection |
Which laws apply to the proposed processing? |
May require particular safeguards or legal arrangements |
| International transfers |
Will personal information be processed across borders? |
May trigger transfer requirements |
| Research ethics |
Does the approved protocol authorize the proposed AI processing? |
May require clarification or amendment |
| Institutional AI policy |
Which tools and tasks are authorized? |
May restrict particular services or information categories |
| Data-sharing agreements |
Who may access and process shared information? |
May limit disclosure to external providers |
| Provider contracts |
What processing, retention, and access terms apply? |
May affect suitability for restricted data |
| Scientific verification |
Who evaluates consequential AI outputs? |
Requires appropriate expertise and evidence |
| Publication requirements |
What AI assistance must be disclosed? |
May require coordinated author declarations |
Not every requirement applies to every project. The relevant obligations should be determined from the actual processing context.
Why One Country's Data-Protection Rules May Not Be Sufficient
International projects may involve several legal frameworks.
For example, the European Union's General Data Protection Regulation establishes requirements for personal data processing and international transfers when it applies.
In the Philippines, Republic Act No. 10173 establishes the national framework for personal information protection.
Other participating jurisdictions may have their own applicable legislation and institutional requirements.
Researchers should not assume that compliance with one country's rules automatically satisfies every other obligation.
Legal applicability depends on the organizations, data subjects, processing activities, and other relevant circumstances.
Institutional data-protection or legal personnel should clarify complex cross-border arrangements.
Why Research Ethics Approvals Must Be Coordinated
International projects may involve research ethics review by more than one institution.
Different committees may approve related procedures subject to different conditions.
For example, an approval may authorize confidential interview analysis by specified researchers without addressing processing through an external generative AI provider.
Introducing such processing may require clarification or an amendment, depending on the applicable protocol.
Researchers should not assume that approval from one institution automatically modifies another institution's conditions.
The team should establish which authority must review proposed changes before they are implemented.
Map Cross-Border AI Data Flows
AI processing may involve more than the visible transfer between a researcher and an application.
Information may pass through provider infrastructure, connected services, subprocessors, or administrative systems.
The precise arrangement depends on the service configuration and contractual terms.
Researchers should identify where restricted information will be submitted, who may process it, and which safeguards apply.
Where cross-border transfer requirements are relevant, the team should obtain appropriate institutional guidance.
Mapping these flows is especially important when collaborators assume that information remains within their university merely because they access AI through an institutional account.
Why AI Provider Policies Require Careful Comparison
Different AI services may offer different retention, access, security, and contractual conditions.
Even the same provider may apply different terms across consumer, institutional, and other deployments.
The team should examine the arrangements applicable to the actual accounts being used.
A commitment not to use submitted data for model training does not automatically resolve retention, security, or confidentiality concerns.
The related problem of different AI privacy policies across research collaborators should be addressed before restricted information is processed.
Should All International Collaborators Use the Same AI System?
Not necessarily.
Using one approved environment may simplify technical support and documentation.
However, different researchers may require different tools for programming, literature exploration, or language assistance.
Some institutions may also prohibit the shared system for particular data categories.
The team should determine whether each proposed use is authorized and scientifically appropriate.
Standardization is useful only when the resulting arrangement satisfies the applicable obligations.
Establish Different Permissions for Different Research Tasks
A practical governance arrangement should distinguish routine assistance from consequential or restricted activities.
| AI activity |
Illustrative governance approach |
| Explaining public research concepts |
Standing permission subject to applicable tool rules |
| Suggesting literature-search terms |
Permitted with independent searching and verification |
| Editing nonconfidential prose |
Conditional permission with appropriate disclosure |
| Generating analytical code |
Testing, documentation, and qualified review |
| Changing primary research methods |
Review by responsible researchers and any required institutional authority |
| Processing confidential participant information |
Specific authorization and appropriate safeguards |
| Generating substantive interpretations |
Independent evidential evaluation and relevant coauthor communication |
| Fabricating research observations |
Prohibited as misrepresentation of research evidence |
The categories should be adapted to the research context rather than adopted as universal institutional rules.
Why Verification Responsibilities Should Be Explicit
AI-generated outputs may influence literature synthesis, analytical procedures, and research interpretation.
International teams often divide responsibilities among collaborators with different expertise.
For example, one researcher may generate statistical code while another evaluates the model's assumptions.
The team should identify who verifies consequential outputs and how the verification will be documented.
AI assistance does not transfer responsibility for scientific accuracy to the system.
The question of coauthor responsibility for verifying AI-assisted work is particularly important when research activities are distributed across institutions.
How Should Teams Handle AI-Generated Literature?
AI may suggest references or summarize publications relevant to the project.
However, generated bibliographic details and claims may be inaccurate.
Researchers should verify original sources before incorporating consequential claims.
This is especially important in international projects where collaborators may work across languages and disciplinary traditions.
A fluent English-language synthesis should not be treated as evidence that the system accurately represented publications in other languages.
The responsible researchers should examine the actual sources and document material limitations.
How Should Teams Handle AI-Assisted Qualitative Analysis?
Qualitative research involves different methodological and epistemological approaches.
AI may support preliminary organization or generate alternative interpretations, but its suitability depends on the selected methodology.
For example, Braun and Clarke (2022) emphasize the researcher's active interpretive role in reflexive thematic analysis.
Automatically accepting generated themes may be inconsistent with the intended analytical approach.
International teams should also consider whether translation and cultural context affect interpretation.
AI-generated summaries may obscure locally meaningful expressions or distinctions.
Qualified researchers familiar with the language and context should evaluate consequential interpretations.
Why Language Differences Matter in International AI Research
Research collaborators may use AI to translate interviews, improve academic writing, or summarize multilingual literature.
These applications can support communication but may introduce interpretive changes.
For example, a culturally specific expression may be translated into a more general English phrase that loses an important distinction.
Such changes may affect qualitative coding or the interpretation of participant experiences.
Researchers should consider the linguistic and cultural expertise needed to verify consequential translations.
Language assistance should not be confused with independent confirmation that the original meaning has been preserved.
How Should International Teams Document AI Use?
Documentation should make consequential AI-assisted procedures understandable across institutions.
A useful record may identify the AI system, task, relevant configuration, material output, verification undertaken, and final human decision.
For computational work, tested code and analytical records may be more useful than an extensive conversation history.
For interpretation, documentation may explain how automated suggestions were evaluated against original evidence.
Records should be stored through authorized arrangements, particularly when they contain confidential information.
Documentation requirements should be proportionate to the research and applicable institutional rules.
How Should AI Disclosure Be Coordinated Across Coauthors?
International collaborators may be familiar with different publisher or disciplinary expectations concerning AI disclosure.
Before submission, the team should identify material AI contributions and review the intended journal's current requirements.
The ICMJE provides recommendations concerning AI-assisted technologies and human author accountability.
Other publication venues may apply different rules.
The corresponding author should obtain accurate information from collaborators before making declarations on behalf of the manuscript.
Internal disclosure should support verification and informed approval, not merely satisfy an administrative requirement.
What If Institutions Disagree About AI-Assisted Authorship?
Authorship should be determined according to the applicable criteria for human scholarly contribution and accountability.
AI use does not automatically establish or eliminate a human collaborator's authorship qualification.
Under ICMJE recommendations, AI-assisted technologies should not be listed as authors because they cannot satisfy human authorship responsibilities.
Other disciplines may use different contribution criteria.
International teams should clarify authorship expectations early, particularly when AI performs technical tasks previously associated with particular contributor roles.
The actual human contribution remains central.
How Should Research Teams Resolve Governance Disagreements?
The collaboration agreement should identify which decisions can be made by designated researchers and which require institutional review.
For example, a disagreement about statistical modeling may require qualified methodological consultation.
A disagreement about cross-border participant data processing may require institutional data-protection guidance.
An authorship dispute may require established institutional or journal procedures.
The principal investigator may coordinate these discussions but should not assume authority to override binding requirements.
Disagreements should be documented accurately and resolved before consequential processing proceeds.
What If Unauthorized AI Processing Occurs?
The team should identify what information was processed, which system received it, and which institutional or legal obligations may have been affected.
Relevant authorities should be informed according to applicable incident-response procedures.
Depending on the circumstances, containment, provider contact, notification, or other corrective action may be necessary.
Deleting an AI conversation does not necessarily establish that all information previously processed has been removed.
The team should also determine whether the incident affects the reliability or permitted use of the research data.
Scientific correction and data-protection compliance are distinct responsibilities.
Why International Collaboration Agreements Should Address AI Explicitly
Traditional research collaboration agreements may cover data access, intellectual property, publication responsibilities, and dispute resolution without specifically addressing generative AI.
AI can introduce new processing arrangements and questions about intellectual contribution.
Teams may therefore benefit from adding task-specific AI provisions.
These provisions should identify authorized processing, verification responsibilities, disclosure expectations, and procedures for introducing new tools.
The question of including AI-use expectations in research collaboration agreements should be considered before substantial shared AI processing begins.
Why Governance Should Be Reviewed Throughout the Project
International research projects may last several years.
During that period, AI services, institutional policies, provider terms, and research activities may change.
An arrangement approved for preliminary literature exploration may not authorize later analysis of confidential participant records.
Similarly, changes to a provider's contractual or technical conditions may require reassessment.
Teams should establish review points before sensitive data processing, consequential analysis, and manuscript submission.
Changes should be documented and communicated before implementation.
Why AI Governance Must Preserve Scientific Accountability
Compliance with data-protection requirements does not automatically establish that AI-generated research outputs are scientifically valid.
Likewise, a correct analytical result does not retroactively authorize inappropriate processing.
Responsible international collaboration requires both governance of information and verification of scientific work.
The European Code of Conduct for Research Integrity emphasizes reliability, honesty, respect, and accountability.
These principles provide a useful foundation for coordinating responsibilities across institutions.
The goal is not to eliminate differences in research practices, but to ensure that consequential AI-assisted decisions remain transparent, defensible, and attributable to responsible human researchers.
