03 · What You Need to Know
How Copyright Similarity Can Arise in AI-Generated Research Images
AI Image Generation Does Not Guarantee Visual Originality
Generative image models produce outputs using patterns learned during training and information supplied through prompts or reference images.
The resulting image may appear new while sharing visual features with existing works.
However, shared features do not necessarily indicate copying of protected expression.
For example, two illustrations may independently depict a researcher standing beside a computer without raising meaningful copyright concerns.
More distinctive similarities involving composition, recognizable characters, or unusual visual arrangements may warrant closer examination.
Researchers should avoid treating the act of generation as proof that an image is free from third-party rights.
Copyright Protects Expression Rather Than General Ideas
Copyright generally protects qualifying original expression rather than abstract concepts or ideas.
For example, the idea of depicting research integrity as a balance between competing considerations is not ordinarily protected merely because an illustrator previously used that metaphor.
However, a particular illustration may contain protectable expressive choices involving arrangement, detailed design, and other original visual elements.
Researchers should distinguish the general visual concept from the specific expression used to communicate it.
This distinction is central to evaluating whether an AI-generated image reproduces protected material.
Artistic Style and Protected Expression Are Different
Researchers may ask AI to create images in a particular artistic style.
For example, a prompt may request a watercolor illustration, three-dimensional cartoon rendering, or minimalist scientific diagram.
General artistic styles are not ordinarily protected by copyright as such in several legal systems.
However, a generated image may still reproduce protected expressive elements associated with an identifiable work.
Researchers should therefore distinguish broad stylistic resemblance from copying of a specific protected composition or character.
Other legal or contractual concerns may arise depending on the circumstances.
Prompting for a Specific Existing Image Increases the Need for Scrutiny
A researcher may request an image that closely resembles an existing journal illustration or textbook diagram.
Such prompts can increase the possibility that the output reproduces distinctive visual elements.
For example, asking AI to recreate a particular published figure with only minor color changes may raise different concerns from requesting a new diagram explaining the same scientific concept.
Researchers should examine whether the output independently communicates the information or reproduces the original work's expressive design.
Changing a few visual details does not automatically eliminate copyright concerns.
Reference Images Can Influence the Result
Some image generation systems allow users to upload photographs, illustrations, or diagrams as references.
The generated output may preserve elements of the supplied image, depending on the requested transformation.
If the reference is copyrighted, researchers should consider whether they are authorized to use it in that manner.
Permission to view or download an image does not necessarily authorize creating and publishing an adaptation.
Researchers should review the applicable license, contractual terms, and statutory exceptions.
AI May Reproduce Distinctive Visual Compositions
Visual similarity can involve more than identical objects.
For example, two illustrations may contain similar subjects but differ substantially in arrangement, perspective, and expressive detail.
Conversely, an AI-generated image may closely reproduce an unusual combination of objects, positioning, and design choices from an existing work.
Researchers should examine the overall relationship between the images rather than relying solely on individual matching elements.
The legal significance of such similarities depends on the jurisdiction and relevant copyright standards.
Recognizable Characters Can Raise Additional Concerns
AI-generated illustrations may resemble distinctive fictional characters from films, comics, or other media.
Some characters may receive copyright protection through their original expression.
Trademark and other legal considerations may also be relevant.
For example, placing a recognizable commercial character in a research poster may raise questions beyond ordinary artistic style.
Researchers should not assume that changing the character's clothing or colors necessarily eliminates the concern.
Using an independently designed character may be more appropriate when the original character is not essential to the research communication.
Scientific Diagrams Can Contain Protected Expression
Research figures often communicate scientific facts, processes, and theoretical relationships.
The underlying scientific information may be distinct from the original visual expression used to present it.
For example, the general concept of illustrating a research workflow through connected stages is not necessarily protected.
However, a particular diagram may include original artwork, distinctive layouts, and expressive design elements.
Researchers should distinguish recreating a scientifically established relationship from reproducing a published diagram's protected appearance.
Accurate Scientific Relationships Do Not Require Identical Artwork
Researchers can often explain the same scientific concept using independently developed visual arrangements.
For example, a conceptual model may be presented through a new diagram that preserves the relationships among variables without copying the original illustration's decorative elements.
However, the scientific meaning must remain accurate.
Changing arrows, labels, or causal relationships merely to make a diagram look different may distort the underlying theory.
Researchers should preserve scientific fidelity while developing appropriate original expression.
AI May Reproduce Elements of Published Research Figures
An AI-generated figure may contain elements resembling charts, diagrams, or illustrations from published articles.
For example, a generated conceptual diagram may reproduce a distinctive layout from a previously published framework.
Researchers should verify whether the original figure contains protected expressive elements and whether reproduction is permitted.
They should also acknowledge the intellectual source of the framework where appropriate.
Attribution and copyright authorization are separate requirements.
AI-Generated Figures Can Be Scientifically Inaccurate
Copyright concerns should not overshadow scientific verification.
Generative AI may produce visually appealing figures containing incorrect labels, misleading arrows, nonexistent structures, or inaccurate relationships.
For example, an AI-generated diagram may reverse the direction of a theoretical relationship while preserving an attractive layout.
Researchers should compare every consequential visual element with the underlying scientific evidence.
A legally permissible image is not necessarily scientifically accurate.
AI Can Introduce Errors During Image Modification
Researchers may use AI to improve the appearance of an existing research figure.
However, image editing can unintentionally alter data points, annotations, scale markings, or diagram relationships.
For example, an automated redesign may reposition an arrow so that it appears to indicate a different causal pathway.
Researchers should verify the modified figure against the original data or conceptual specification.
Visual improvement should not change the scientific meaning without explicit justification.
Photorealistic AI Images Can Create Additional Risks
AI-generated photographs may resemble identifiable individuals, locations, or existing photographs.
Copyright is only one possible concern.
Privacy, personality rights, consent, and misleading representation may also matter.
For example, a synthetic image of a laboratory should not be presented as documentary evidence of an experiment that was never conducted.
Researchers should clearly distinguish illustrative imagery from empirical documentation.
AI-Generated Images Should Not Replace Actual Research Evidence
Illustrative graphics and empirical figures serve different purposes.
An AI-generated illustration may help communicate a research concept.
However, it should not be presented as a photograph of actual participants, laboratory conditions, or observed results unless it genuinely documents those circumstances.
Researchers should preserve the distinction between generated visual representations and recorded evidence.
Misrepresenting synthetic images as empirical observations can raise serious research integrity concerns independently of copyright.
Similarity Detection for Images Has Limitations
Reverse image searches and visual similarity systems may help identify existing images resembling an AI-generated output.
However, these tools do not provide complete coverage of copyrighted works.
They may also identify visually similar images that do not share protected expression.
Researchers should inspect relevant matches rather than treat automated similarity scores as legal determinations.
The absence of a matching result does not guarantee that an image is original.
Visual Similarity Should Be Evaluated Contextually
Researchers should consider the type of image, distinctive features, and nature of the overlap.
For example, two ordinary bar charts using similar colors may share conventional design choices.
Two elaborate scientific illustrations reproducing an unusual composition may present different considerations.
The legal analysis depends on the protected elements and applicable law.
Researchers should avoid assuming that a fixed percentage of visual similarity determines infringement.
Changing Colors Does Not Automatically Avoid Infringement
Researchers may attempt to avoid similarity by changing the palette, cropping the image, or replacing a few objects.
However, these changes may leave substantial protected expression intact.
For example, an illustration may retain the same distinctive composition and arrangement despite using different colors.
Researchers should examine whether the final image independently expresses the concept rather than merely disguising the original.
Attribution Does Not Automatically Authorize Reproduction
Citing the creator of an existing image may satisfy scholarly attribution requirements.
However, citation alone does not necessarily authorize reproduction or adaptation of protected artwork.
Researchers should examine relevant licenses, permissions, and copyright exceptions.
For example, a publisher may require permission to reproduce a figure even when the source is fully cited.
Academic acknowledgment and legal authorization should be evaluated separately.
Open-Access Images May Still Be Copyrighted
Images available in open-access articles may remain subject to copyright.
Some are distributed under Creative Commons licenses that permit specified forms of reuse.
However, the license applied to an article may not necessarily cover every third-party image included within it.
Researchers should inspect figure credits and relevant licensing information.
They should not assume that every image visible in an open-access publication can be reused without conditions.
Creative Commons Licenses Have Different Conditions
Creative Commons licenses provide different permissions concerning attribution, commercial use, and adaptations.
For example, some licenses permit modification under specified conditions, while others restrict distribution of adapted material.
Researchers should identify the license attached to the actual source image.
They should also distinguish material covered by the license from excluded third-party content.
Compliance depends on the particular license and intended use.
Public-Domain Images Require Appropriate Verification
Some images may be in the public domain because copyright protection has expired, does not apply, or has been relinquished where legally possible.
However, public-domain status can vary across jurisdictions.
Researchers should verify the relevant status rather than rely on an unsupported website label.
Scholarly attribution may remain appropriate even when copyright permission is unnecessary.
AI Provider Terms May Affect Publication Rights
Image generation services may establish terms concerning output use, commercial exploitation, uploaded reference material, and third-party rights.
These terms vary among providers and account arrangements.
Researchers should review the agreement applicable to the service used.
Contractual permission does not necessarily establish copyright protection or eliminate third-party claims.
Copyright Ownership of AI Images Is a Separate Question
A researcher may be permitted to use an AI-generated image without necessarily possessing exclusive copyright in it.
In jurisdictions requiring human authorship, purely AI-generated visual expression may not qualify for protection.
Original human modifications, selection, or arrangement may receive protection under applicable law.
Researchers should distinguish ownership of qualifying expression from permission to publish the generated image.
Publisher Policies May Restrict AI-Generated Figures
Scholarly publishers may impose requirements concerning AI-generated images, graphical abstracts, research figures, and disclosure.
Some policies distinguish illustrative artwork from figures representing scientific data.
Researchers should consult the current policy of the intended journal.
Permission from an AI provider does not automatically establish compliance with publisher requirements.
Conference Presentations May Also Require Appropriate Rights
Researchers sometimes assume that copyright considerations apply only to journal publications.
However, conference posters, presentation slides, institutional websites, and publicly distributed reports can also involve reproduction or distribution of protected material.
The applicable requirements depend on the use and jurisdiction.
Researchers should evaluate rights before public dissemination rather than only at manuscript submission.
Philippine Copyright Law Requires Local Consideration
In the Philippines, Republic Act No. 8293, the Intellectual Property Code, as amended, provides the principal copyright framework.
Researchers should examine relevant provisions concerning protected artistic works, economic rights, and limitations.
Foreign copyright guidance may provide useful background but should not automatically be treated as controlling Philippine law.
Where substantial rights are involved, qualified legal advice may be appropriate.
Copyright and Trademark Concerns Should Be Distinguished
AI-generated images may include recognizable logos, symbols, or branding.
Some of these elements may raise trademark concerns independently of copyright.
For example, including a recognizable publisher logo in a research poster may create questions about endorsement or authorized use.
Researchers should evaluate the intended presentation and applicable rights.
Removing a copyright concern does not necessarily resolve trademark or misleading endorsement issues.
Researchers Should Preserve Image Provenance
Documentation may include the AI service, generation date, prompts, uploaded references, editing steps, and final output.
Researchers should also retain relevant source comparisons and permissions.
These records can help explain how the image was produced.
However, provenance documentation does not independently establish that the image is legally unrestricted.
Its value lies in supporting transparency and verification.
Researchers Should Distinguish Illustration From Scientific Evidence
AI-generated images can support research communication when their purpose is clearly illustrative.
However, researchers should not allow visual appeal to substitute for accurate representation.
For example, an AI-generated illustration of an experimental procedure should not imply that the depicted equipment or process was actually used unless that is accurate.
Scientific claims must remain grounded in verified evidence.
Independent Visual Design May Reduce Some Copyright Risks
Researchers can develop original graphics from verified concepts, data, and design specifications.
For example, a conceptual diagram can be constructed from independently established relationships rather than by closely imitating an existing illustration.
However, independent creation does not automatically resolve every legal concern, particularly when protected source material is incorporated.
Researchers should evaluate both provenance and the final visual expression.
Copyright Review Should Be Proportionate to the Intended Use
A generic decorative illustration may present different concerns from a detailed graphic resembling an identifiable copyrighted work.
Researchers should prioritize examination of distinctive similarities, uploaded references, recognizable characters, and substantial adaptations.
Where uncertainty is consequential, permission, replacement, or qualified legal advice may be appropriate.
The objective is not to treat every resemblance as infringement but to avoid publishing material without examining credible rights concerns.
