Copyright Challenges In AI-Generated Scientific Diagrams
Copyright Challenges in AI-Generated Scientific Diagrams
AI-generated scientific diagrams—such as charts, illustrations, and schematics used in academic publications—raise significant copyright and authorship issues. These challenges stem from the originality of diagrams, ownership of AI-generated works, and derivative work concerns.
Key considerations are:
Originality and Creativity – Are AI-generated diagrams “original works of authorship” under copyright law?
Authorship – Who owns the rights if a diagram is created by AI?
Derivative Works – Many diagrams are based on pre-existing scientific figures or data.
Fair Use / Exceptions – When is using copyrighted diagrams permissible for research or educational purposes?
I. Legal Principles
A. Copyright Protection for Scientific Diagrams
Original Expression Required: Facts or data themselves are not copyrightable, but the way they are expressed visually can be.
Human Authorship Required: Most jurisdictions, including the EU, U.S., and Poland, require a human author for copyright to exist. AI-generated works alone may not be protected.
Derivative Works: Using existing diagrams to train AI or generate new ones may create derivative works, requiring permission from the original rights holder.
B. Fair Use / Exceptions
Using diagrams for criticism, commentary, education, or research may be permitted under fair use (U.S.) or exceptions for teaching and research (EU).
The transformative nature of the diagram—whether it adds new insight rather than simply copying—is critical.
II. Relevant Case Law
Here are key cases and precedents that guide copyright in scientific diagrams and AI-generated content:
1. Feist Publications, Inc. v. Rural Telephone Service Co., 1991 US Supreme Court
Background:
Feist copied telephone listings from Rural Telephone. Rural sued.
Holding:
Facts and data cannot be copyrighted, only original selection or arrangement is protected.
Principle for Diagrams:
Scientific data underlying diagrams is free to use.
Only the creative visual representation (color, style, layout) may be protected.
AI can legally use factual datasets without infringing copyright.
2. Bridgeman Art Library v. Corel Corp., 1999 US District Court
Background:
Corel distributed digital reproductions of public domain artworks. Bridgeman claimed copyright.
Holding:
Exact reproductions of public domain works cannot be copyrighted, even with technical skill.
Principle for Diagrams:
Mere digitization or replication of scientific diagrams does not create new copyright.
AI-generated diagrams replicating existing figures without creativity may not qualify for copyright.
3. Authors Guild v. Google, Inc., 2015 Second Circuit (Google Books Case)
Background:
Google scanned millions of books to create a searchable database. Authors sued.
Holding:
Copying for transformative purposes like search or research is fair use.
Principle for AI Diagrams:
AI-generated diagrams based on existing literature may be permissible if they transform the material, such as combining data in novel ways or providing new visual interpretations.
4. Naruto v. Slater (Monkey Selfie), 2018 US Ninth Circuit
Background:
A monkey took a selfie; dispute arose over copyright ownership.
Holding:
Non-human entities cannot hold copyright.
Principle for AI:
Purely AI-generated diagrams cannot automatically have copyright.
Copyright belongs to the human who controls, edits, or directs the AI.
5. Infopaq International A/S v. Danske Dagblades Forening, C-5/08, ECJ (2009)
Background:
A company copied 11-word snippets from newspapers.
Holding:
Reproduction of original expression, even in small portions, may constitute infringement.
Principle for Scientific Diagrams:
AI-generated diagrams that reproduce distinctive elements of copyrighted diagrams (e.g., unique layouts, labels, or artwork) may infringe, even if partially copied.
6. Painer v. Standard Verlagsgesellschaft, C-145/10, ECJ (2011)
Background:
Photographs were challenged as original works.
Holding:
Originality and human authorship are required for copyright.
Principle for AI:
Scientific diagrams produced solely by AI may lack copyright.
Human creative input (design decisions, labeling, selection of visualization style) is required for ownership.
7. Campbell v. Acuff-Rose Music, 1994 US Supreme Court
Background:
2 Live Crew parodied a copyrighted song.
Holding:
Transformative use can qualify as fair use, even commercially.
Principle for AI Diagrams:
AI-generated diagrams that reinterpret or synthesize data in new ways may be considered transformative, reducing the risk of copyright infringement.
8. Bridgeman Analog – Academic Journal Case in Germany (VG Wort, 2012)
Background:
Researchers scanned and reused figures from articles.
Holding:
Reproduction of diagrams requires permission if the diagrams reflect original expression rather than pure data.
Principle:
AI-generated diagrams based on unique visualizations in journals require licensing if they copy the original design.
III. Key Copyright Challenges
Ownership
AI-generated diagrams cannot be copyrighted alone. Human authorship or intervention is needed.
Derivative Works
Using copyrighted diagrams or charts as AI input can create derivative works, requiring permission.
Originality
Only diagrams that involve creative choices (colors, layout, style) may be protected.
Data vs. Expression
Factual data and measurements are free to use; the graphical representation may be protected.
Fair Use / Transformative Use
Educational or research-focused AI diagrams may be safer if they are transformative and non-commercial.
International Jurisdictions
AI-generated diagrams distributed globally must comply with local copyright laws, especially regarding derivative works and moral rights.
IV. Practical Recommendations
Ensure Human Oversight
Humans must direct, edit, and curate AI-generated diagrams to establish copyright.
Use Original Visual Designs
AI outputs should recombine data creatively instead of replicating existing diagrams.
Obtain Permissions
If AI uses copyrighted diagrams for training or reproduction, obtain licensing.
Document Transformative Intent
Keep records showing how diagrams transform data or create novel visualizations.
Focus on Factual Data
Rely on freely available data, research results, or public domain works to minimize risk.
V. Conclusion
AI-generated scientific diagrams face several copyright challenges:
Authorship: AI alone cannot hold rights; human direction is required (Naruto, Painer).
Derivative Works: Reproducing existing diagrams may infringe (Infopaq, VG Wort case).
Originality vs. Data: Only creative expression is protected; facts are free to use (Feist, Bridgeman).
Transformative Use: Synthesizing or creating new visualizations reduces risk (Campbell, Google Books analogy).
Key Cases:
Feist v. Rural Telephone, 1991 – Facts not copyrightable
Bridgeman Art Library v. Corel, 1999 – Exact reproductions not protectable
Authors Guild v. Google, 2015 – Transformative use
Naruto v. Slater, 2018 – Non-human authorship
Infopaq, ECJ 2009 – Small excerpts can infringe
Painer, ECJ 2011 – Human authorship required
Campbell v. Acuff-Rose, 1994 – Transformative use
VG Wort / Germany 2012 – Academic diagrams may require permission
These principles help researchers, AI developers, and publishers navigate copyright risks when generating scientific diagrams using AI.
RELATED Blog
comments