Copyright OwnershIP For AI-Curated Digital Art Collections.

06 Mar 2026 --
0 Comments

📌 1. Thaler v. Perlmutter (U.S. — Human Authorship Required)

Summary:
Stephen Thaler sought copyright protection in the U.S. for a visual artwork entitled “A Recent Entrance to Paradise,” which was created autonomously by his AI system called DABUS.

Legal Issue:

Can a work generated entirely by an AI qualify for copyright?

Do AI systems have legal authorship?

Court Ruling:

The U.S. Copyright Office rejected the application, affirming that copyright requires human authorship.

Federal courts upheld the Office’s ruling, and the U.S. Supreme Court declined to review the case in March 2026, leaving the human‑authorship requirement intact.

Key Legal Principles:

Copyrightable works must be created by a natural person, not a machine.

AI‑generated works lacking meaningful human creative input are ineligible for protection under U.S. law.

Relevance:
This case is foundational in the U.S. framework, essentially precluding autonomous AI from obtaining copyrights in its outputs and limiting ownership to humans who exercise creative direction over the AI.

📌 2. Li v. Liu (Beijing Internet Court, China — AI‑Assisted Work Can Be Copyrightable)

Summary:
Mr. Li generated AI images using Stable Diffusion and posted them online. When another blogger used one of the images without authorization, Li sued for copyright infringement.

Legal Issue:

Can an AI‑generated work be copyrightable if a human provided creative guidance and intellectual input?

Court Ruling:

The Beijing Internet Court ruled that the AI‑generated image was copyrightable because Mr. Li provided detailed prompts, selected parameters, and repeatedly refined outputs, showing his intellectual achievement and personalised expression.

The AI itself was not regarded as an author; instead, human input was considered sufficient to confer copyright.

Key Legal Principles:

Chinese copyright law requires originality and intellectual creation rather than strictly “human authorship” as in the U.S.

Human selection and creative decisions in crafting the final AI output were decisive in granting copyright.

Relevance:
This case represents a more flexible legal approach where creative human engagement with AI can produce a copyrightable work, even if a machine generated the final image.

📌 3. Visual Artists’ Class Action (Stability AI, Midjourney, DeviantArt) (U.S. — Infringement & Training Data)

Summary:
A group of artists sued AI companies alleging their copyrighted images were used without permission to train generative image models like Stable Diffusion and Midjourney.

Legal Issue:

Does using copyrighted images to train AI models constitute infringement?

Are the images generated by those models unlawful reproductions or derivative works?

Case Status & Rationale:

The case raises complex questions about whether AI training on copyrighted content qualifies as copying or an infringing derivative.

Some claims argued the models could “reproduce” or mimic the plaintiffs’ works too closely, diluting artists’ market opportunities.

Key Legal Principles:

Courts also assess how the AI processes input data — whether it merely stores it or transforms it in non‑infringing ways.

Relevance:
This ongoing litigation illustrates how courts may treat data used for training and the outputs of generative AI, which is central to digital art collections curated or generated by AI.

📌 4. GEMA v. OpenAI (Munich Regional Court, Germany — AI Reproduction of Lyrics)

Summary:
Germany’s music rights society GEMA sued OpenAI, alleging that the company’s training of ChatGPT models on copyrighted song lyrics enabled the AI to reproduce those lyrics without authorization.

Legal Issue:

Can an AI model infringe copyright by memorizing and reproducing protected content it was trained on?

Court Ruling:

The Munich court held that the AI did violate copyright when it memorised and subsequently reproduced significant portions of protected lyrics, viewing such output as a form of reproduction under German and EU law.

Key Legal Principles:

An AI system’s output that directly reflects copyrighted materials in its training dataset can be infringement.

Courts may treat such reproduction akin to making unauthorized copies.

Relevance:
This case marks an important precedent in Europe for treating AI training data use and output reproduction as subject to copyright protection.

📌 5. Suryast – RAGHAV AI Painting Case (India/Canada/US) (Mixed Outcomes)

Summary:
Artist‑lawyer Ankit Sahni created an AI‑generated artwork named Suryast using the RAGHAV AI painting app and sought copyright registration in various countries.

Legal Issues:

Who qualifies as author or co‑author when AI is involved?

Can AI‑generated works be copyrighted at all?

Outcomes:

In India, the copyright office initially granted registration with the AI tool listed as a co‑author, then later questioned whether AI can be an author due to lack of legal personality.

In the United States, the Register refused copyright registration, saying there was insufficient human creativity to support authorship.

Canada reportedly granted protection recognizing the human author’s role.

Key Legal Principles:

Different jurisdictions apply different standards for human involvement, originality, and authorship in AI contexts.

Canada’s approach allowed co‑authorship with AI to be recognised, highlighting significant global divergence.

Relevance:
Suryast demonstrates how AI‑art copyright ownership decisions vary widely, especially regarding what constitutes sufficient human input.

📌 6. Zarya of the Dawn (U.S. Copyright Office Practice)

Summary:
Though not a full court case, the U.S. Copyright Office dealt with the application for a comic book with AI‑generated art.

Legal Issue:

Does AI‑generated imagery in a larger creative work qualify for copyright?

Outcome:

The Office ultimately limited the copyright to the text and compilation elements, excluding purely AI‑generated images, emphasising that only human‑authored content can be protected.

Key Legal Principles:

Compilation copyrights are available for human creative contributions, even when the work contains AI components.

AI‑generated portions may remain unprotected if lacking human creative input.

Relevance:
This administrative decision reinforces the human‑authorship standard within U.S. copyright practice and explains how AI components in mixed works are treated.

📌 Key Legal Themes & Ownership Takeaways

🧠 1. Human Authorship Is Central Yet Interpreted Differently

U.S.: Strictly requires human authorship for copyright eligibility. AI alone cannot be an author.

China: Allows AI‑generated works to be copyrightable if the human provides sufficient creative input, such as prompt selection and refinement.

📊 2. Training Data and Infringement

European courts (e.g., GEMA v. OpenAI) may treat AI reproduction of copyrighted training content as infringing.

🖼 3. Mixed & Collaborative Works

Works incorporating AI‑generated elements may receive partial copyright protection for human‑contributed parts.

🌍 4. International Divergence

The global legal landscape is not uniform — jurisdictions vary widely from U.S. reluctance to recognise AI involvement to China’s acceptance of human‑guided AI outputs as copyrightable.

📌 Key Legal Principles in AI Art Copyright

Principle	U.S.	China	Europe	India/Canada
Human Authorship Required?	Yes	Flexible (human creative input)	Varies by case, copyrightable outputs possible	Mixed outcomes depending on human input
AI as an Author?	No	No	Generally no	No (India), Canada co‑authorship recognized
Ownership of Training Output	Depends, may be infringement	Depends	Can be infringement	Depends

📌 Practical Implications for AI‑Curated Art Collections

AI‑generated collections may have copyright only if human curation involves meaningful creative choices — selecting, arranging, modifying, or editing outputs to reflect a unique artistic vision.

Ownership rights depend heavily on jurisdiction and level of human involvement.

E.g. projects in China may secure stronger rights for AI‑assisted works, while in the U.S. they might not qualify at all.

Collections that merely compile or reproduce existing works without creative arrangement typically lack copyright (similar to compilations in Feist‑type principles).

Legal frameworks are rapidly evolving, with courts continuing to refine originality and authorship standards in the age of generative AI.