03 · What You Need to Know
Why Lawful Access and AI Processing Permissions Are Different
Lawful Access Establishes How You Obtained the Publication
Lawful access generally means that a researcher obtained or accessed a publication through an authorized route.
Examples include institutional subscriptions, individual purchases, authorized repositories, and open-access publisher platforms.
However, the permissions associated with these arrangements may differ.
For example, a university subscription may allow authorized users to read and download articles for scholarly purposes while imposing conditions on automated processing or redistribution.
Researchers should identify the actual source and terms of their access.
Lawful acquisition is relevant to copyright analysis, but it does not independently determine every subsequent permitted use.
Access Rights and Reproduction Rights Are Distinct
Copyright systems commonly recognize rights associated with reproducing protected works.
Accessing a publication does not necessarily transfer those rights to the reader.
For example, a researcher may be permitted to download an article for personal academic study without acquiring unrestricted permission to distribute copies to external organizations.
Uploading the document to an AI platform may involve reproductions or transmissions that require separate consideration.
The legal significance depends on the applicable law and actual processing arrangement.
Uploading a PDF May Create Additional Copies
Cloud-based AI services commonly process uploaded files through computing infrastructure.
Depending on the system, this may involve transmission, temporary storage, document extraction, indexing, or retention.
These operations can differ from opening the PDF on a personal device.
Researchers should determine what happens to uploaded content under the applicable service terms.
However, the creation of a technical copy does not automatically establish infringement. Relevant authorization and statutory exceptions must also be considered.
Institutional Subscriptions Are Governed by Agreements
Universities frequently negotiate access to scholarly publications through institutional licensing agreements.
These agreements may define authorized users, permitted uses, downloading conditions, automated analysis rights, and restrictions on third-party sharing.
For example, an agreement may authorize specified forms of text and data mining while imposing conditions on how the content is processed.
Researchers should not assume that all publishers or subscriptions apply identical restrictions.
The university library may be the appropriate starting point for clarifying the applicable agreement.
Personal Article Purchases Do Not Necessarily Grant Unlimited Rights
Purchasing a research paper does not ordinarily mean acquiring the copyright in the work.
The purchase may provide access under specified conditions.
For example, a researcher may be permitted to read and retain a copy without being authorized to redistribute it publicly.
Whether uploading it to a particular AI service is permitted depends on the purchase terms and applicable law.
Researchers should distinguish ownership of a copy from ownership of copyright.
Open Access Does Not Always Mean Unrestricted Use
Open-access publications are available through arrangements that may permit reading and reuse.
However, open-access licenses differ in the permissions they grant.
For example, Creative Commons Attribution licenses generally permit copying and adaptation subject to applicable conditions.
Other licenses may contain noncommercial or no-derivatives restrictions.
Researchers should examine the license associated with the specific article.
The term open access alone does not establish that every form of AI processing is authorized.
Creative Commons Licenses Require Specific Interpretation
Creative Commons licenses distinguish among different reuse permissions and obligations.
For example, CC BY permits broad reuse subject to attribution and other relevant conditions.
CC BY-NC restricts uses that fall within the license's definition of commercial purposes.
Licenses containing ND provisions restrict sharing adapted material, but do not necessarily prohibit every private modification or analytical activity.
Researchers should determine which licensed rights are exercised by the actual AI workflow.
They should avoid treating all license conditions as blanket prohibitions on computational processing.
Third-Party Content May Have Different Rights
A research article may contain photographs, figures, tables, or illustrations reproduced from other sources.
These components may be excluded from the article's general reuse license.
For example, an open-access paper may contain a photograph reproduced with permission from a separate rights holder.
Researchers should examine figure credits and licensing notices when the AI task involves reproducing or extracting such material.
Permission to process the article's main text may not automatically cover every embedded component.
Text and Data Mining May Be Covered by Specific Permissions
Text and data mining involves computational analysis of text or other materials to identify information or patterns.
Some jurisdictions recognize statutory exceptions for particular text and data mining activities.
Institutional licenses may also authorize specified forms of computational analysis.
However, these permissions can have conditions concerning lawful access, eligible organizations, retained copies, and the purposes of processing.
Researchers should determine whether their intended AI workflow falls within the relevant authorization.
AI Summarization and Text Mining Are Not Necessarily Identical
Uploading an article to a conversational AI system may involve document storage, retrieval, summarization, and generated responses.
These activities may differ from conventional text and data mining workflows.
For example, a service that retains uploaded documents or reproduces extensive passages may present considerations not addressed by a narrowly defined mining permission.
Researchers should examine the actual operations performed rather than assume that every form of AI processing is legally equivalent.
Statutory Exceptions Can Affect the Analysis
Copyright exceptions may permit particular uses of protected material without obtaining separate permission.
These provisions differ across jurisdictions.
For example, the European Union's Directive 2019/790 establishes provisions concerning text and data mining, subject to specified conditions.
In the United States, fair use involves a contextual assessment of statutory factors.
Researchers should identify the law applicable to the proposed activity.
Foreign exceptions should not automatically be treated as applicable to research conducted elsewhere.
Research Purpose Does Not Automatically Establish Fair Use
Researchers sometimes assume that any activity undertaken for academic purposes is automatically permitted.
However, copyright exceptions may require consideration of several factors.
For example, the amount of material used, the purpose of processing, and the effect on relevant markets may matter under certain legal frameworks.
Researchers should not assume that noncommercial research is always exempt.
Nor should they assume that every computational use requires separate permission when an applicable exception may authorize it.
Publisher Contracts and Copyright Exceptions Must Be Distinguished
A publisher may impose contractual conditions concerning the use of licensed content.
Statutory copyright exceptions operate under a different legal framework.
Whether a particular contractual restriction is enforceable against an otherwise permitted statutory use depends on the jurisdiction.
Researchers should avoid assuming that publisher terms automatically override every legal exception.
They should also avoid assuming that an exception necessarily eliminates contractual obligations.
Consequential conflicts may require qualified legal interpretation.
The AI Provider's Data Practices Matter
AI services differ in how they process, retain, and potentially reuse uploaded content.
Some services may offer organizational controls that differ from consumer arrangements.
Researchers should review the terms applicable to their actual service and account.
Relevant considerations include retention, model improvement, subprocessors, deletion, and access controls.
These conditions can affect confidentiality and contractual compliance.
However, a service's privacy protections do not independently establish copyright permission.
Private AI Conversations Are Not Automatically Legally Authorized
A researcher may believe that uploading a paper is acceptable because the AI conversation is not publicly visible.
However, private processing may still involve transmission or reproduction of protected material.
Whether those activities are permitted depends on applicable rights, exceptions, and agreements.
Researchers should not confuse interface privacy with legal authorization.
Local AI Processing May Reduce Some Risks
Running a model locally may avoid transmitting a publication to an external service.
This can reduce certain confidentiality and third-party disclosure concerns.
However, local processing may still involve copying or transforming copyrighted material.
Researchers should examine the relevant rights regardless of whether processing occurs locally or remotely.
Technical architecture affects the risk assessment but does not independently determine legality.
Institutionally Approved AI Tools May Offer Different Arrangements
Some universities provide AI services through organizational agreements.
These arrangements may establish additional security controls or clarify permitted processing activities.
However, institutional approval of an AI platform does not automatically authorize uploading every copyrighted publication.
Researchers should confirm whether the relevant content license permits the intended use.
Approval of the tool and authorization to process particular material are separate questions.
Publisher-Integrated AI Features May Have Specific Permissions
Some scholarly platforms provide AI-assisted search, summarization, or analysis within their own services.
These features may operate under agreements that differ from those governing uploads to unrelated AI providers.
Researchers should examine the scope of the platform's authorized functionality.
However, integration does not necessarily mean that all content can be exported or redistributed without restriction.
Confidential Manuscripts Require Additional Consideration
Researchers may receive unpublished manuscripts through peer review, collaboration, or institutional duties.
These materials may be subject to confidentiality obligations independently of copyright.
For example, a peer reviewer should not assume that uploading a confidential manuscript to an external AI service is permitted.
The relevant journal or institutional policy should be consulted.
Where authorization is absent, the material should not be disclosed.
Peer Review Access Is Not General Processing Authorization
Peer reviewers receive manuscripts for a specific scholarly purpose.
That access may be subject to explicit confidentiality requirements.
Some publishers restrict or prohibit uploading confidential submissions to generative AI systems.
Researchers should verify the applicable policy rather than rely on general copyright assumptions.
Permission to evaluate a manuscript does not necessarily authorize transferring it to a third-party service.
Personal Information Creates Separate Legal Obligations
Research documents may contain identifiable participant information or sensitive institutional records.
Uploading such materials may raise data protection and ethics concerns.
Researchers should evaluate applicable lawful processing grounds, consent arrangements, security requirements, and institutional policies.
Copyright authorization does not eliminate these obligations.
Philippine Copyright Law Requires Local Analysis
In the Philippines, Republic Act No. 8293, the Intellectual Property Code, as amended, provides the principal statutory copyright framework.
Its provisions concerning economic rights and limitations may be relevant to computational processing of scholarly publications.
However, the legality of a particular AI upload cannot be established solely from the fact that the researcher accessed the article through a Philippine university.
The applicable contractual arrangements and processing circumstances also matter.
Qualified advice may be appropriate when the legal consequences are substantial.
Institutional Library Staff Can Clarify Access Agreements
Researchers may not have direct access to the complete licensing agreements negotiated by their universities.
Library personnel responsible for electronic resources may be able to identify the applicable terms.
For example, they may clarify whether external text and data mining is permitted under a particular subscription.
Researchers should seek institutional clarification rather than infer permissions from the presence of a download button.
Permission May Differ Across Versions of the Same Article
A study may be available as a preprint, accepted manuscript, or final publisher version.
These versions can have different distribution and reuse conditions.
For example, an author-posted manuscript may carry a different license from the publisher's formatted PDF.
Researchers should examine the rights associated with the specific version they intend to process.
They should not assume that permission concerning one version automatically applies to every version.
Open Repositories May Provide Authorized Alternatives
Institutional and disciplinary repositories may contain versions of publications available under more permissive terms.
Researchers may be able to use these versions when their licenses permit the intended processing.
However, repository availability alone does not establish unrestricted reuse rights.
The relevant license or authorization should still be checked.
Researcher-Prepared Notes May Reduce Unnecessary Uploading
Researchers can independently read articles and prepare structured notes describing methods, findings, and limitations.
AI may then assist with organizing or comparing those notes.
This approach can reduce the need to transmit complete copyrighted documents.
However, notes containing extensive copied expression may still raise copyright questions.
Researchers should preserve accurate attribution and avoid reproducing protected material without an appropriate basis.
Bibliographic Metadata May Be Sufficient for Some Tasks
Some AI-assisted research tasks do not require full-text access.
For example, generating search strategies, organizing verified references, or preparing preliminary screening criteria may be possible using bibliographic metadata.
Researchers should provide only the material necessary for the intended task.
However, metadata-based assistance should not be presented as a verified analysis of the complete article.
Authorized Processing Does Not Guarantee Accurate Summaries
Even when an AI upload is legally permitted, the generated analysis may contain errors.
For example, AI may misstate sample sizes, omit methodological limitations, or attribute findings incorrectly.
Researchers should compare consequential summaries with the original publications.
Legal authorization and scientific reliability are separate requirements.
AI May Reproduce Protected Passages in Its Response
An AI assistant may reproduce or closely paraphrase text from an uploaded article.
Whether that output may be published or redistributed requires separate consideration.
For example, authorization for private computational analysis does not necessarily authorize public reproduction of substantial protected passages.
Researchers should distinguish permission to process source material from permission to disseminate generated content.
Bulk Uploading Can Introduce Additional Conditions
Processing hundreds or thousands of journal articles may involve different contractual and technical considerations from analyzing a single document.
Some agreements may contain specific provisions concerning automated extraction or large-scale computational use.
Researchers should examine the scope and scale of the intended workflow.
Permission for ordinary downloading should not automatically be interpreted as authorization for bulk processing.
International Collaboration Can Complicate Permissions
Research teams may include collaborators from institutions with different subscriptions and licensing arrangements.
One researcher's lawful access does not necessarily authorize distribution of the PDF to every collaborator or shared AI workspace.
Researchers should verify relevant sharing and processing rights.
Where appropriate, collaborators can use their own authorized access routes or properly licensed versions.
AI Service Terms Can Change
Providers may revise their data retention policies, processing arrangements, or contractual conditions.
Researchers should review the terms applicable when the workflow is conducted.
They should not rely indefinitely on an earlier understanding of the service.
Institutional governance procedures should account for material changes affecting approved research workflows.
Researchers Should Document Their Authorization
Where AI-assisted document processing materially contributes to a research project, researchers should retain sufficient documentation of the relevant permissions.
This may include the publication license, institutional guidance, service terms, and processing purpose.
Such records can support transparency and accountability.
However, documentation alone does not make an otherwise unauthorized activity lawful.
Permission Should Be Evaluated Before Uploading
Researchers should establish the relevant authorization before transmitting protected documents.
If permission remains uncertain, they may seek clarification or use an authorized alternative.
The fact that an AI service technically accepts a PDF does not establish that the researcher has the legal or contractual right to upload it.
The most defensible workflow combines verified permissions, appropriate data handling, and independent assessment of the resulting scientific content.
