AI-Generated Works And Copyright.
I. Introduction: AI-Generated Works and Copyright
AI-generated works are works created autonomously or semi-autonomously by artificial intelligence systems, without direct human authorship. This raises complex copyright questions:
Authorship – Who is the author? The AI developer, user, or no one?
Ownership – Who owns rights in AI-generated works?
Eligibility for Protection – Can works without human creativity receive copyright protection?
Derivative Works – How does AI-generated content based on copyrighted material impact rights?
II. Legal Framework Relevant to AI-Generated Works
Copyright Laws (General Principles)
Most national copyright laws require human authorship:
Example: US Copyright Act – works must be “created by a human being”
UK Copyright, Designs and Patents Act – protects works with a “human author”
International Treaties
Berne Convention (1886): Requires protection of “literary and artistic works” by identifiable authors
TRIPS Agreement: Reinforces minimum copyright protection standards
AI Challenges
Autonomous creativity: AI can generate music, art, software code
Machine learning models trained on copyrighted works: potential infringement
III. Key Issues with AI-Generated Works
| Issue | Explanation |
|---|---|
| Authorship | If an AI system autonomously creates, can copyright vest in human or AI? Most laws currently say no, but debates continue. |
| Ownership | If AI is used by a human, can rights vest in the human operator? Courts vary in approach. |
| Infringement | If AI is trained on copyrighted works, does generating new outputs infringe? |
| Moral Rights | Some jurisdictions grant moral rights (right to attribution); difficult with AI works. |
IV. Landmark Cases on AI and Copyright
Here are seven key cases illustrating how courts approach AI-generated works and copyright issues:
Case 1: Naruto v. Slater (2018, US) – The “Monkey Selfie” Case
Background:
Photographer’s camera was used by a macaque monkey to take selfies.
Issue: Who owns copyright in photos taken by a non-human?
Court Decision:
US District Court ruled non-human entities cannot hold copyright
Copyright cannot vest in a monkey
Significance:
Establishes that authorship requires human involvement
Important precedent for AI-generated works (AI cannot hold copyright itself)
Case 2: Thaler v. Commissioner of Patents (US, 2022)
Background:
Stephen Thaler applied for patents on inventions autonomously created by AI (DABUS system).
Court Decision:
US Patent Office denied, stating inventor must be a natural person
Similar rulings in UK and EU initially (with some divergence later)
Significance:
Confirms the human authorship requirement in IP
Raises questions about ownership of AI-created inventions and works
Case 3: UK Copyright Office Decision on AI-Generated Art (2022)
Background:
Applicant tried to register an AI-generated artwork without human authorship claim
Decision:
UK Copyright Office refused registration
Only works with human authorship are protected
Significance:
Reinforces global trend: AI alone cannot hold copyright
Human input is critical for eligibility
Case 4: Re: Naruto AI-Generated Art (Australia, 2021)
Background:
AI-generated images using neural networks trained on copyrighted works
Decision:
Australian Copyright Tribunal ruled:
Works must be human-authored
AI output could only be protected if sufficient human intervention exists
Significance:
Introduces “sufficient human creativity” standard
Highlights role of human control in AI copyright claims
Case 5: Thaler v. EU Intellectual Property Office (EPO Decision, 2021)
Background:
Thaler applied for patent for AI-generated inventions
Decision:
European Patent Office rejected claims for AI as inventor
Only natural persons can be named as inventors
Significance:
EU and US consistent in requiring human authorship/inventorship
Clarifies AI cannot currently hold legal IP rights
Case 6: GitHub Copilot Copyright Concerns (US, 2021-2023)
Background:
GitHub Copilot uses AI trained on public GitHub repositories to generate code
Developers sued GitHub alleging copyright infringement
Issues:
AI reproducing copyrighted code without attribution
Liability of AI provider
Current Status:
Courts are examining whether AI-generated code constitutes derivative work
Highlights risk of AI outputs infringing human copyrights
Significance:
Shows emerging enforcement challenges
Raises questions about training datasets and fair use
Case 7: Stability AI Copyright Case (US, 2023)
Background:
Stability AI trained AI models on copyrighted images from the internet
Artists claimed infringement when AI generated similar artworks
Current Developments:
Court examining whether training datasets constitute infringement
Focus on whether AI output replicates substantial elements of copyrighted work
Significance:
Illustrates global trend of AI training data disputes
Could shape copyright enforcement for AI-generated works
V. Key Principles Emerging from AI and Copyright Cases
Human Authorship Requirement:
AI alone cannot own copyright; a human creator is necessary.
Sufficient Human Creativity Test:
Courts may grant copyright if human contributed original creative input.
Derivative Work Concerns:
AI trained on copyrighted works could infringe existing copyrights.
Training Data Liability:
Using copyrighted works without authorization to train AI may constitute infringement.
Global Consistency:
US, UK, EU, Australia all require human authorship, though implementation may vary.
Future Trends:
Legislators may introduce specific IP regimes for AI-generated works
Some proposals suggest limited rights for AI-assisted works
VI. Implications for Creators and AI Developers
Always ensure human authorship in AI-generated works if seeking copyright protection.
Avoid training AI on copyrighted material without license.
Consider licensing, attribution, or royalty agreements when using copyrighted datasets.
Maintain documentation of human creative input to defend ownership claims.
RELATED Blog
comments