03 · What You Need to Know
Why Generative AI Can Misinterpret Local Research Governance
Research Governance Differs Across Countries
Research systems operate within national legal frameworks, institutional policies, disciplinary standards, and professional requirements.
For example, procedures governing human participant research, personal information, institutional data access, and research funding may differ across jurisdictions.
An AI-generated recommendation may describe a legitimate requirement in one country while incorrectly presenting it as applicable elsewhere.
Researchers should therefore distinguish general research ethics principles from specific legal or administrative obligations.
The existence of an internationally recognized practice does not establish that every institution implements it through identical procedures.
AI May Treat Familiar Research Systems as Universal
Generative AI may produce recommendations based on institutional arrangements frequently described in its available information.
For example, it may refer to an Institutional Review Board using terminology associated with particular national research systems.
Other institutions may use research ethics committees, ethics review boards, or differently organized approval mechanisms.
The terminology itself is not necessarily incorrect, but assuming that the associated procedures apply universally can be misleading.
Researchers should confirm the actual review structure and approval requirements of their institution.
Institutional Policies May Be More Specific Than National Requirements
National legislation establishes important obligations, but institutions may also maintain policies governing research administration, data access, ethics review, and publication.
For example, a university may require internal authorization before researchers access student records or recruit participants through institutional communication channels.
These procedures may not be described in publicly available national legislation.
An AI system may therefore provide a legally relevant general explanation while omitting institution-specific requirements.
Researchers should consult the appropriate institutional offices and current policy documents.
AI Can Confuse Research Ethics Approval With Administrative Permission
Research ethics approval and administrative authorization serve different purposes.
Ethics review generally evaluates relevant participant protections, research risks, and ethical considerations.
Administrative authorization may concern access to institutional facilities, records, personnel, or communication systems.
Depending on the study, both may be required, or different arrangements may apply.
Researchers should not assume that obtaining one form of approval automatically satisfies every other requirement.
AI-generated checklists should be verified against the institution's actual governance procedures.
AI May Misinterpret Philippine Data Protection Requirements
In the Philippines, Republic Act No. 10173, known as the Data Privacy Act of 2012, provides an important legal framework for personal data protection.
However, determining the obligations applicable to a particular research activity requires examining the relevant provisions, implementing rules, lawful processing grounds, and institutional arrangements.
An AI system may incorrectly suggest that all research data processing requires the same form of consent or that research activities are automatically exempt from privacy obligations.
Neither generalization should be accepted without verification.
Researchers should consult applicable guidance from the National Privacy Commission and their institution's data protection personnel where appropriate.
Consent and Lawful Data Processing Are Related but Distinct
Research ethics consent and legal grounds for processing personal information are not necessarily identical concepts.
For example, an ethics review may require informed consent as part of participant protection, while applicable privacy legislation may recognize different lawful grounds for particular processing activities.
The relationship depends on the jurisdiction and research circumstances.
AI-generated advice may incorrectly merge these concepts or assume that one automatically satisfies the other.
Researchers should evaluate both ethical obligations and applicable data protection requirements.
AI Can Invent Institutional Policies
Generative AI may produce plausible names for offices, committees, forms, or approval procedures that do not exist.
For example, a model might state that a particular university requires approval from a named research committee without having access to the institution's current policies.
The recommendation may sound credible because similar committees exist elsewhere.
Researchers should verify the existence and authority of every consequential institutional requirement.
A confident AI response should not be treated as evidence that an office or policy exists.
Outdated Information Can Produce Incorrect Compliance Advice
Research policies and regulations may change over time.
Institutions may revise ethics procedures, data governance policies, funding requirements, or administrative responsibilities.
AI systems may rely on information that does not reflect the latest changes.
Even retrieval-enabled systems may encounter outdated webpages or superseded policy documents.
Researchers should verify the effective date, current status, and applicability of relevant requirements.
Where authoritative sources conflict, clarification should be sought from the responsible office rather than allowing AI to select a convenient interpretation.
AI May Misidentify the Responsible Regulatory Authority
Different agencies have different mandates.
For example, educational regulation, personal data protection, research funding, and professional licensing may involve distinct authorities.
An AI system may incorrectly attribute one agency's responsibilities to another.
Researchers should verify which organization has jurisdiction over the relevant activity.
They should also distinguish mandatory legal requirements from nonbinding guidance or institutional recommendations.
Accurate identification of the responsible authority is necessary for reliable compliance decisions.
International Guidelines Are Not Automatically Binding Local Law
International research ethics guidelines can provide valuable principles for responsible research.
However, their legal status and institutional adoption vary.
For example, a recommendation from an international organization may be influential without automatically creating a legal obligation in every country.
Researchers should determine whether a guideline is binding, formally adopted, incorporated into institutional policy, or offered as professional guidance.
AI systems may fail to distinguish these categories.
Researchers should avoid describing a voluntary recommendation as a statutory requirement without an appropriate legal basis.
Local Research Funding Rules Can Be Misunderstood
Research grants may impose specific eligibility criteria, reporting obligations, expenditure restrictions, and publication requirements.
These conditions vary among funding organizations and programs.
An AI system may provide generic grant-management advice that does not match the actual funding agreement.
For example, it may incorrectly assume that particular equipment purchases or personnel expenses are automatically allowable.
Researchers should consult the relevant funding guidelines and signed agreements.
AI can help organize grant requirements but should not independently determine compliance.
AI May Misunderstand Institutional Research Roles
Universities differ in how they distribute responsibility among research offices, ethics committees, academic departments, data protection personnel, and institutional administrators.
An AI-generated workflow may assign responsibilities to the wrong office.
For example, it may assume that the office responsible for research grants also authorizes access to student records.
Researchers should confirm the actual institutional division of responsibilities.
Accurate administrative mapping can prevent unnecessary delays and inappropriate disclosure of research information.
Cross-Border Research Introduces Additional Complexity
Research involving participants, collaborators, or data systems across countries may raise multiple legal and institutional requirements.
For example, a multinational study may involve different rules concerning personal data transfers, ethics review, and institutional authorization.
AI may oversimplify these arrangements by applying the requirements of only one jurisdiction.
Researchers should identify the relevant locations of participants, institutions, data processing, and storage.
Appropriate legal and institutional guidance may be necessary when obligations overlap.
AI Can Misinterpret Local Educational Structures
Educational institutions use different program classifications, academic calendars, assessment structures, and delivery arrangements.
For example, a fully asynchronous online program should not automatically be treated as equivalent to a conventional course delivered through videoconferencing.
If an AI system assumes scheduled classroom attendance is required, its methodological recommendations may become inappropriate.
Researchers should describe the actual institutional context and verify the system's understanding.
Incorrect assumptions about educational delivery can affect sampling, measurement, and interpretation.
Local Terminology Can Create Ambiguity
Institutional terms may have specific meanings within particular countries or organizations.
For example, a term describing academic standing, program accreditation, or research authorization may not correspond directly to similarly named processes elsewhere.
AI may translate unfamiliar terminology into a familiar but inaccurate category.
Researchers should preserve official terminology when necessary and provide contextual explanations.
Where a term has legal or administrative significance, its meaning should be verified through the relevant authoritative source.
AI May Give Correct General Advice for the Wrong Reason
An AI system may recommend an action that is appropriate while providing an inaccurate explanation of why it is required.
For example, seeking ethics review may be prudent, but the AI may incorrectly cite a law or institutional policy as the source of the obligation.
Researchers should verify both the recommendation and its stated authority.
A correct practical suggestion does not validate an incorrect legal explanation.
This distinction matters when researchers document compliance or communicate requirements to collaborators.
AI-Generated Regulatory Citations Require Verification
Generative AI can invent legal provisions, misquote policies, or cite superseded documents.
Researchers should verify the existence, wording, effective date, and applicability of cited requirements.
Official government websites, institutional policy repositories, and authenticated regulatory publications should be preferred where available.
Secondary summaries may help with orientation but should not replace the controlling source for consequential decisions.
Researchers should not reproduce unverified legal citations in protocols, ethics applications, or institutional communications.
How Can Researchers Evaluate AI Advice About Local Regulations?
Researchers can establish a set of jurisdiction-specific questions with answers verified through authoritative sources.
They may then compare AI-generated responses against those reference materials.
Relevant evaluation criteria include factual accuracy, identification of the responsible authority, recognition of exceptions, and appropriate expression of uncertainty.
Repeated tests may reveal recurring misunderstandings.
However, regulatory evaluation should be updated when relevant laws or policies change.
The results should be interpreted within the jurisdictions and questions actually examined.
Local Expertise Can Identify Errors That Generic AI Advice Misses
Institutional research officers, ethics personnel, data protection specialists, and relevant legal professionals may possess information unavailable to an AI system.
For example, they may know whether a particular approval form has been replaced or whether a new administrative procedure applies.
Researchers should consult the appropriate office when consequential requirements are uncertain.
However, informal advice should also be distinguished from official institutional determinations.
Where possible, consequential decisions should be supported by current written guidance.
Providing Context Can Improve AI Responses but Does Not Guarantee Accuracy
Researchers can specify the country, institution, research population, data types, and intended procedures when requesting AI assistance.
This may reduce inappropriate assumptions.
For example, identifying a study as involving adult university students in the Philippines may produce more relevant preliminary questions than a prompt that simply asks about human participant research.
However, additional context does not guarantee that the AI knows current local requirements.
Researchers should still verify consequential claims independently.
Researchers Should Maintain a Compliance Verification Record
When AI assists with regulatory or institutional planning, researchers should record which recommendations were verified and through which sources.
Relevant documentation may include the official policy title, issuing authority, effective date, applicable provision, and institutional clarification.
Researchers should distinguish confirmed requirements from unresolved questions.
This record can help explain why particular procedures were adopted.
It also reduces the risk that an unverified AI-generated statement becomes embedded in subsequent research documents.
AI Should Support Compliance Research Rather Than Make Compliance Decisions
Generative AI can help researchers formulate questions, identify possible obligations, summarize supplied policy documents, and organize administrative tasks.
However, its role should remain subordinate to authoritative requirements and appropriate institutional judgment.
Where the consequences of error are substantial, researchers should obtain qualified clarification before proceeding.
The objective is not to avoid AI entirely but to prevent plausible language from replacing verified regulatory knowledge.
