IP Rights In AI-Generated Vietnamese Calligraphy Styles.
1. IP RIGHTS IN AI-GENERATED VIETNAMESE CALLIGRAPHY
AI systems generating Vietnamese calligraphy (Thư pháp Việt Nam) typically involve:
1. Algorithms / Models
Deep learning models (GANs, diffusion models) that produce stylized calligraphy or mimic historical masters.
Protected as:
Patents – if the AI method or image-generation algorithm is novel and non-obvious.
Trade secrets – proprietary model architecture, weights, or training pipelines.
2. Training Data
Datasets include images of historical calligraphy, scanned manuscripts, or annotated character sets.
Facts (character shapes, text content) are not copyrightable.
Curated datasets or collections of annotated calligraphy can be protected under copyright or database rights.
3. Software Implementation
The code powering AI generation, web apps, or desktop tools.
Copyrightable as software.
Proprietary modules (e.g., style transfer layers) may be trade secrets.
4. Outputs / Generated Art
AI-generated Vietnamese calligraphy images or stylized text files.
Copyrightability depends on whether the output shows creative, human-like expression.
Courts increasingly examine whether AI output can be assigned to a human creator for protection.
Key IP Questions
Can AI-generated calligraphy hold copyright?
Who owns the rights: AI developer, user, or model trainer?
Are training datasets protected?
Can trade secrets in AI models prevent transparency?
2. KEY CASE LAWS AND THEIR APPLICATION
1. Feist Publications, Inc. v. Rural Telephone Service Co. (1991)
Facts:
Feist used factual phone listings without permission.
Holding:
Facts themselves are not copyrightable, only creative selection or arrangement is.
Relevance to AI Calligraphy:
Raw Vietnamese characters and historical calligraphy images are facts; they cannot be monopolized.
Curated, annotated, or uniquely arranged calligraphy datasets may receive copyright.
Principle:
IP protects creative compilation, not raw factual data.
2. Diamond v. Diehr (1981)
Facts:
Patent on a rubber curing process using a computer algorithm.
Holding:
Software-assisted processes are patentable if tied to a practical, physical process.
Relevance:
AI generating Vietnamese calligraphy for physical reproduction (posters, prints, signage) can be considered applied to a tangible outcome, making it potentially patentable.
Principle:
Practical application of AI may support patent protection.
3. Alice Corp. v. CLS Bank (2014)
Facts:
Patent claimed abstract computer-implemented ideas.
Holding:
Abstract ideas are not patentable unless they include inventive concepts.
Relevance:
AI algorithms for style transfer or calligraphy synthesis must show novel approaches beyond standard neural networks to be patentable.
Principle:
Generic AI methods alone are insufficient for patent protection.
4. Naruto v. Slater (2018)
Facts:
“Monkey selfie” case where AI/animal took a photograph.
Holding:
Only humans can hold copyright; animals (or AI) cannot.
Relevance:
AI-generated calligraphy cannot hold copyright by itself.
Rights must be assigned to a human developer, user, or organization.
Principle:
AI cannot be an author under current copyright law.
5. State v. Loomis (2016)
Facts:
COMPAS algorithm challenged in sentencing due to secrecy.
Holding:
Trade secrets are valid, but courts cannot rely solely on opaque AI.
Relevance:
AI models generating Vietnamese calligraphy may be proprietary, but users may demand transparency for commercial or artistic use, licensing, or verification.
Principle:
Trade secrets cannot override accountability and explainability.
6. International News Service v. Associated Press (INS) (1918)
Facts:
INS copied AP news for redistribution.
Holding:
Recognized quasi-property rights in time-sensitive, commercially valuable information.
Relevance:
AI-generated calligraphy outputs can have commercial or cultural value.
Vendors or artists could assert limited proprietary rights over generated pieces.
Principle:
Valuable outputs, even from public datasets, may have limited protection.
7. DABUS AI Inventor Cases (2019–2025)
Facts:
AI system listed as inventor in patents.
Holding:
Courts rejected AI as inventor; only humans/entities may hold IP.
Relevance:
AI-generated calligraphy cannot hold IP; rights must be assigned to human developers, artists, or institutions.
Principle:
Ownership of AI-generated works resides in humans or organizations.
3. SYNTHESIS OF LEGAL PRINCIPLES
| Component | IP Protection | Legal Principle |
|---|---|---|
| Algorithms / Models | Patent / Trade Secret | Must be novel and practically applied (Diamond, Alice) |
| Training Data | Limited copyright / Database rights | Facts public, curated datasets protectable (Feist) |
| Software Implementation | Copyright / Trade Secret | Protects code and proprietary modules |
| AI-Generated Calligraphy Outputs | Quasi-property / Copyright if human-assigned | Commercially valuable outputs may have limited protection (INS, Naruto) |
| AI Ownership | N/A | AI cannot hold IP; rights must be human/organization-owned (DABUS) |
| Transparency / Accountability | N/A | Trade secrets cannot override verification/explainability (Loomis) |
4. PRACTICAL CHALLENGES
Authorship Assignment: Who owns copyright for AI calligraphy?
Vendor Lock-In: Proprietary AI models may limit licensing.
Cultural Sensitivity: Vietnamese heritage works may invoke moral rights.
Derivative Works: AI may replicate existing styles, raising infringement risks.
Licensing & Monetization: Commercialization requires clarity of IP ownership.
5. CONCLUSION
AI-generated Vietnamese calligraphy exists in a hybrid IP landscape:
Algorithms: Patentable or trade secret
Training Data: Public facts; curated datasets protectable
Software: Copyrightable
Outputs: Quasi-property or copyright if human-assigned
AI itself: Cannot hold IP
Legal trend:
While IP rights protect AI innovations commercially, transparency, explainability, and assignment to humans are critical, especially for culturally sensitive or artistic works.
RELATED Blog
comments