1. Legal Nature of Neural Art

Neural art can involve three layers:

(A) Input layer (training data)

• images, text, music, cultural datasets
• may include copyrighted works

(B) Model layer (algorithm)

• neural network architecture
• weights, parameters, code

(C) Output layer (generated art)

• images, videos, sound, interactive art

2. Core Legal Problem

3. Major Case Laws & Decisions (More than 5 detailed cases)

CASE 1: Thaler v. United States Copyright Office (US AI authorship case)

Facts:

Stephen Thaler attempted to register a copyright for an artwork created entirely by an AI system (“Creativity Machine”).

He listed the AI as the sole author.

Legal Issue:

Can a non-human system be recognized as an author under copyright law?

Decision:

• US courts rejected copyright registration
• Held that:
  • Copyright requires human authorship
  • AI-generated works without human creative input are not protected

Significance:

• Foundational case for AI-generated art
• Establishes human authorship doctrine

CASE 2: Naruto v. Slater (monkey selfie principle extended to AI debates)

Facts:

A monkey took photographs using a camera owned by a photographer (David Slater).

Legal Issue:

Can a non-human entity own copyright?

Decision:

• Court held:
  • Animals cannot hold copyright
  • Copyright requires human authorship

Relevance to Neural Art:

• Used as analogical authority in AI cases
• Reinforces principle that:
  • non-human creators cannot be authors

Significance:

• Frequently cited in AI art disputes worldwide

CASE 3: Zarya of the Dawn Copyright Registration (US Copyright Office AI ruling)

Facts:

A graphic novel was created using:

• human-written prompts
• AI-generated illustrations (Midjourney-like system)

Legal Issue:

Which parts of AI-assisted art are copyrightable?

Decision:

• Text and arrangement: protected
• AI-generated images: not protected
• Only human-authored creative selection qualifies

Significance:

• Introduced “modicum of human creativity test”
• Established split ownership model

CASE 4: Getty Images v. Stability AI (AI training data dispute)

Facts:

Getty Images sued AI developers for allegedly using copyrighted images to train neural networks without permission.

Legal Issue:

Does training an AI model on copyrighted images constitute:

• reproduction
• infringement
• unauthorized copying?

Decision (procedural stage, ongoing in many jurisdictions):

• Courts allowed parts of the claim to proceed
• Central issue: whether model training = copying in legal sense

Significance:

• Landmark dispute on data ingestion rights
• Could define legality of neural art systems

CASE 5: Andersen v. Stability AI / Midjourney (US class action neural art case)

Facts:

Artists alleged that generative AI systems:

• trained on billions of copyrighted images
• created derivative works without consent

Legal Issues:

• copyright infringement via training
• derivative work generation
• moral rights violations

Outcome:

• Case ongoing (no final ruling yet)
• Court acknowledged serious legal questions regarding:
  • data scraping
  • transformative use doctrine

Significance:

• One of the most important cases shaping AI art law
• Could redefine fair use vs infringement boundaries

CASE 6: UK Intellectual Property Office Guidance Cases on AI-generated works

Facts:

The UK considered whether AI-generated works qualify under the Copyright, Designs and Patents framework.

Legal Issue:

Who is the author of computer-generated works?

Outcome:

UK law currently states:

• author = person who makes “arrangements necessary for creation”
• AI output may be protected if human involvement exists

Significance:

• UK is one of few jurisdictions allowing non-traditional authorship attribution
• Provides a hybrid protection model for neural art

CASE 7: Tencent AI Art Ownership Dispute (China AI copyright cases)

Facts:

Tencent AI-generated images were disputed between:

• platform developers
• users who generated prompts

Legal Issue:

Who owns AI-generated content:

• user or platform?

Decision (Chinese court trend):

• If human demonstrates:
  • selection
  • input creativity
  • iterative control
    → user may be recognized as author

Significance:

• China adopts user-centric authorship approach
• More flexible than US strict human authorship rule

CASE 8: Nvidia GAN-generated art ownership disputes (industry practice cases)

Facts:

GAN (Generative Adversarial Network) systems produce artworks used in commercial design pipelines.

Disputes arise over:

• ownership of outputs
• licensing of model-generated assets

Legal Issue:

Are outputs:

• derivative works of training data?
• or new independent works?

Outcome:

• No uniform global ruling
• Companies rely on:
  • licensing agreements
  • contractual ownership clauses

Significance:

• Shows shift from IP law → contract law dominance in AI art

CASE 9: EU Copyright Directive Interpretation on AI outputs

Facts:

European courts and policymakers considered whether AI-generated works qualify for copyright.

Legal Issue:

Originality requirement in EU law.

Outcome:

• EU stance:
  • copyright requires “author’s own intellectual creation”
  • AI-only output usually not protected
  • human creative input required

Significance:

• Strong rejection of non-human authorship in EU system

CASE 10: Indian AI-generated art copyright advisory practice (policy approach cases)

Facts:

Indian copyright authorities and advisory interpretations address AI-generated works in emerging digital art sectors.

Legal Issue:

Whether AI outputs can be registered under existing law.

Outcome:

• Generally:
  • human involvement required for protection
  • pure AI output uncertain or unprotected

Significance:

• Reflects cautious approach similar to US/EU but still evolving

4. Key Legal Principles from Neural Art Cases

(1) Human Authorship Rule

Most jurisdictions require:

• human creativity
• human control over expression

(2) AI as Tool, Not Author

AI is treated like:

• camera
• paintbrush
• design software

(3) Split Protection Model

• Human contributions → protected
• AI-generated parts → often unprotected

(4) Training Data Controversy

Ongoing debate:

• whether ingestion = copying
• whether output is derivative

(5) Contract Law Dominance

Because IP law is unclear:

• ownership is often decided by:
  • platform terms
  • licensing agreements

5. Major Legal Challenges

• No global uniform rule for AI authorship
• Massive copyright litigation risk in training data
• Difficulty defining originality in probabilistic systems
• Cross-border inconsistency (US vs China vs EU)
• Moral rights and attribution gaps

Conclusion

Protection of intangible knowledge represented through digital neural art reveals a major transformation in intellectual property law. Across jurisdictions, courts consistently struggle with one central issue: law is built for human creativity, not machine-generated expression.

The current global pattern is:

• Human input = protected
• Pure AI output = largely unprotected
• Training data = legally contested space

Neural art is pushing IP law toward a hybrid future where copyright, contract law, and data governance merge into a new regulatory framework for machine creativity.