Copyright In AI-Generated Vietnamese Epic Ballads
📌 1. Core Principle: Human Authorship Required for Copyright
Legal Rule
Under copyright law in most jurisdictions, including Vietnam, only works created by human intellectual labor are protected. AI cannot be an author.
Vietnamese Context:
Vietnam’s Intellectual Property Law (amended 2022) defines an “author” as a natural person who directly creates the work through their intellectual labor. Works created entirely by AI without human creative input may be objected to for registration because they lack human originality and identifiable author.
This has direct relevance to epic ballads (trường ca):
If an AI simply synthesizes lyrics and melody without meaningful choices or edits by a person, Vietnamese law arguably will not grant copyright.
📌 2. AI Training Data and Copyright: GEMA v. OpenAI (Germany, 2025)
Facts
In GEMA v. OpenAI, the Munich Regional Court ruled that the training of ChatGPT on copyrighted song lyrics (without license) violated copyright law. The court held that AI models that can reproduce copyrighted material due to training effectively infringe if they store/provide that material.
Reasoning
Copyright protected works were used without permission.
The AI’s outputs showed reproduction of lyrics, constituting “copying” of protected material.
Relevance
For Vietnamese epic ballads trained on copyrighted Vietnamese folk songs, using such copyrighted corpora without permission could constitute infringement under similar legal reasoning.
📌 3. Suno & Udio Settlements with Major Music Labels (U.S., 2025)
Facts
Major labels Warner Music Group and others sued AI music generators like Suno and Udio for training AI on copyrighted music without authorization. They settled and agreed to license arrangements.
Why It Matters
Although not a decisive court ruling, these settlements demonstrate legal risk for AI platforms that generate music (including ballads):
Using copyrighted recordings for model training without licenses is legally and commercially risky.
Labels negotiated paid licensing frameworks, effectively treating AI‑generated music as derivative.
For Vietnamese epic music, training AI on existing Vietnamese songs without consent may similarly invite legal action and demands for licensing.
📌 4. Zarya of the Dawn Copyright Revocation (U.S., 2023)
Case Summary
An AI‑illustrated comic (“Zarya of the Dawn”) was initially granted copyright by the U.S. Copyright Office, but later revoked when it was revealed images were generated by AI.
Legal Reasoning
Copyright requires human authorship.
Works entirely produced by AI were held ineligible for protection.
Only human‑created textual and arrangement elements remained protected.
Relevance
If an AI creates the music and lyrics of an epic ballad without human editing, a copyright office might refuse registration — as in the comic case.
📌 5. Li v. Liu (Beijing Internet Court, 2023) – Human Contribution Recognized
Facts
The Beijing Internet Court held in Li v. Liu that a work generated with substantial human creative decision‑making (prompts, parameters, selection, refinement) has copyright protection, even if AI tools were used.
Legal Insight
This shows a middle ground:
AI as a tool → human creative labor is still central.
If a composer provides detailed prompts, edits, selects, and shapes outputs, that human creative imprint may make the work copyrightable.
Application
For Vietnamese epic ballads:
If a composer writes prompts, refines melodies/lyrics, they may claim copyright in their humanly selected and edited creative result.
📌 6. U.S. Copyright Office Policies & Thaler Case (U.S.)
Principle
U.S. copyright doctrine holds that:
AI generated works without human authorship cannot be copyrighted.
Registration requires human creativity.
Thaler v. Perlmutter (U.S. example)
A famous case (often cited in U.S. discussions) involved Stephen Thaler’s attempt to register an AI artwork. The Copyright Office rejected it because it lacked human authorship.
Key Rule: AI output alone = no copyright.
📌 7. European Union Originality Standard (Infopaq)
Fact
The Infopaq International case (CJEU) emphasised that originality requires the author’s own intellectual creation.
Legal Value
Although about text excerpt copying, it reinforces the principle adopted across EU: no copyright unless expression reflects human creativity.
🔎 How This Applies to Vietnamese Epic Ballads
A. Purely AI‑Generated Ballads
If an AI has autonomously composed the lyrics and melody:
Under current Vietnamese practice, such works may be refused registration because they lack human authorship and identifiable creator.
Many foreign jurisdictions reject copyright for pure AI outputs.
B. AI‑Assisted but Human‑Shaped Ballads
If a human composer:
Crafts detailed prompts;
Selects and edits AI outputs;
Adds creative modification;
Then that human may be recognized as the author of the humanly shaped creative work — like in Li v. Liu.
C. Training Data Issues
If the AI was trained using copyrighted Vietnamese folk songs or other protected material without permission:
The AI platform might be legally liable (as seen in GEMA v OpenAI and Suno/Udio disputes).
📌 8. Practical Takeaways for Creators
âś” Document Human Creative Input:
Write, edit, fine‑tune AI outputs extensively — this is essential to claim copyright.
âś” Avoid Unauthorized Training Data:
Using AI trained on protected songs without disposal rights can lead to infringement claims.
âś” Consider Licensing:
As with Suno/Udio, licensing arrangements with rights holders may be necessary for commercial exploitation.
âś” Copyright Registration May Be Denied:
If an application cannot identify a human author, authorities may refuse protection.
📌 9. Summary Table of Key Cases
| Case / Principle | Jurisdiction | Rule |
|---|---|---|
| GEMA v OpenAI | Germany | AI training on copyrighted content can infringe |
| Suno/Udio Settlements | U.S. | AI must license training data to avoid infringement |
| Zarya of the Dawn | U.S. | Pure AI art is ineligible for copyright |
| Li v Liu | China | AI-assisted works with human creative input may be protected |
| U.S. Copyright Office Policy | U.S. | Human authorship required |
| Infopaq Standard | EU | Originality = human intellectual creation |
Conclusion
In short, Vietnam and international law largely converge on this: AI‑generated works by themselves are not copyrightable. However, if significant human creativity shapes the output, those human contributions can qualify for copyright protection. Moreover, training and output that incorporate copyrighted source material without permission — even if transformed — may lead to copyright infringement claims, as seen in cases like GEMA v. OpenAI and the Suno/Udio disputes.

comments