Copyright In Generative AI Comics Created Via Natural Language Prompts.
📌 Fundamental Legal Framework: Human Authorship & AI
Before diving into cases, it helps to understand the core legal question most courts are wrestling with:
Copyright law traditionally requires human authorship — machines, programs, or algorithms cannot be authors under most national laws. This applies to illustrations, text, comics, videos, etc.
Many jurisdictions (U.S., UK, India, China, etc.) are actively debating where AI‑generated works fit, but as of early 2026, most systems do not grant copyright to works solely generated by AI. Instead, they ask:
Did a human meaningfully contribute creativity?
Is the output merely a mechanical execution of prompts?
Does the work infringe third‑party copyrights?
📘 Case 1 — Zarya of the Dawn (U.S. Copyright Office)
Facts:
Author Kris Kashtanova wrote a comic with text, and used Midjourney AI to create all illustrations.
She registered the comic with the U.S. Copyright Office, but did not initially disclose the use of AI.
Ruling:
The Copyright Office confirmed the copyright registration, but later revoked protection for the AI‑generated images because they were not human authored.
The copyright remains valid only for:
➤ The text/story
➤ The selection, arrangement, and compilation of elements
but not the raw AI illustrations.
Why important: This is the first high‑profile test of AI art and comics. It shows that:
✔ Human text + layout + art curation can be copyrightable,
✘ Pure AI images by themselves are not — unless significantly edited by a human.
📘 Case 2 — U.S. Court of Appeals (Thaler / DABUS) — No Copyright for Pure AI Output
Facts:
Inventor Stephen Thaler claimed copyright for an artwork created entirely by his AI system (no human contribution).
Ruling:
The U.S. Court of Appeals held that works “without human involvement” are not eligible for copyright.
The Copyright Act requires human authorship — machines cannot hold rights.
Impact on comics: If a generative AI produces both text and art without meaningful creative choices by a human—such works cannot be copyrighted under U.S. law.
📘 Case 3 — GEMA v. OpenAI (Germany, 2025)
Facts:
GEMA (German music rights body) sued OpenAI, alleging that the AI model memorised and reproduced copyrighted songs without a license in training.
Ruling:
The Regional Court of Munich ruled that AI models can infringe when they memorise and reproduce copyrighted material.
Reproduction in AI outputs can constitute copyright violation under German/EU law.
Takeaway for comics: If an AI comic generator reproduces copyrighted characters or large copyrighted text/art, then it may infringe even if ownership of the AI output can’t be claimed.
📘 *Case 4 — Andersen, McKernan, Ortiz v. Midjourney/Stability/DeviantArt (USA)
Facts:
Three professional artists sued popular AI image platforms (including Midjourney) claiming the systems were trained on millions of copyrighted images without consent.
Status:
Initial filings were amended because early complaints lacked enough detail, but courts allowed some claims to proceed.
Plaintiffs argue the AI models create outputs that are not sufficiently transformative and are too similar to their original copyrighted art.
Relevance to comics: Many generative comic systems use similar training methods, so this case tests whether the training data itself infringes copyrights.
📘 Case 5 — Stability AI & Getty Images (international)
Facts:
Getty Images sued Stable Diffusion’s developer for using millions of its images without a licence to train the generative model.
Status:
Some claims were dismissed for procedural reasons, but Getty pivoted to other legal theories (e.g., trademark/brand misuse and facilitating infringement).
Lesson:
Even when AI outputs can’t be copyrighted, the training process itself can be directly challenged as infringement.
🔍 International Comparison Cases (Legal Doctrines)
🏛️ Li v. Liu (China, Beijing Internet Court)
In China, a court held that if a human issues prompts, selects outputs, modifies parameters, and refines results repeatedly (creative effort), then copyright may attach to the final image.
This reflects a global trend: copyright depends on meaningful human artistic judgment – not mere prompting.
🧠 Key Legal Principles Emerging in AI Comics
1. Human Creative Contribution Required
Most courts (U.S., S. Korea, China, Europe) require substantial human input — not just a prompt — to warrant copyright.
2. AI Output Alone Is Not Copyrightable
Generative AI comics purely created from prompts are often deemed non‑copyrightable.
3. Human Elements (Text, Editing, Arrangement) Can Be Protected
Even if images aren’t protected, the creative arrangement, story, and human edits may still be copyrighted.
4. Third‑Party Copyright Infringement Can Still Occur
AI output can infringe another person’s copyright even if the output itself is uncopyrighted.
🧠 What This Means for AI‑Generated Comics
| Situation | Copyrightable? |
|---|---|
| Comic art generated solely by AI from prompts | ❌ No (U.S., most jurisdictions) |
| Comic text written by a human | ✔ Yes |
| Comic art edited or significantly altered by a human | ✔ Sometimes |
| Comic that reproduces existing copyrighted characters (like superheroes) | ❌ Likely infringement |
| Training a comic AI on copyrighted works without licence | ❌ Potential infringement |
⚖️ Conclusion: Summary Rules
Human authors in AI comics still hold rights, but usually only for the parts they contributed creatively.
Pure generative output is not automatically protected — courts have consistently required human creative input.
Using copyrighted content without a licence is infringement, both in training and in the generated output.
Courts globally are still developing the law — expect evolving decisions.
RELATED Blog
comments