Copyright Control Of AI-Assisted Musical Composition Tools.
I. Core Copyright Issues in AI-Assisted Music Composition
AI-assisted music composition tools typically involve:
Software algorithms – machine learning models or generative AI that produce music
Input data – existing music used for training the AI
Generated output – new music tracks, melodies, harmonies, or arrangements
Key copyright questions include:
Who owns copyright in music generated by AI?
Is the AI tool itself protected as software?
Does using copyrighted music to train AI constitute infringement?
Can derivative works be claimed if AI outputs resemble original music?
1. Copyright in Musical Works
In most jurisdictions (US, EU, Philippines, UK): copyright protects musical compositions, including melody, harmony, rhythm, lyrics, and sometimes arrangement.
Protection applies to expression of ideas, not ideas themselves.
Implication: If an AI reproduces an existing melody or chord progression, it may infringe copyright. Original outputs may be protected depending on jurisdiction.
2. Copyright in AI Software
AI tools themselves are software under copyright law.
Copying source code without authorization → infringement.
Algorithms or underlying machine learning models are not protected, only their expression in code.
3. Derivative Works
Musical derivatives include remixes, adaptations, or arrangements.
AI output may be considered a derivative work if it closely imitates or transforms copyrighted music.
4. Ownership and Authorship
US Copyright Office guidance: works entirely generated by AI without human authorship cannot be copyrighted.
Human input or selection of AI outputs can establish authorship.
Moral rights (in EU) protect the integrity of musical works.
5. Fair Use Considerations
Use of copyrighted music as AI training data may be lawful under fair use in US (research, transformative use).
EU and UK do not have the same explicit exception, requiring licenses or copyright clearance.
II. Relevant Case Laws
Here are more than five detailed cases shaping copyright in AI music tools:
1. Computer Associates Int’l, Inc. v. Altai, Inc. (1992, USA)
Computer Associates International, Inc. v. Altai, Inc.
Court:
United States Court of Appeals for the Second Circuit
Facts:
Altai copied software scheduling code.
Holding:
Developed the Abstraction-Filtration-Comparison (AFC) test
Only expression of code is protected; algorithms and ideas are not
Relevance:
AI music tools: copying AI algorithms is lawful if independently implemented
Copying source code of commercial AI music software → infringement
2. Feist Publications v. Rural Telephone Service (1991, USA)
Feist Publications, Inc. v. Rural Telephone Service Co.
Facts:
Telephone directory listings claimed as protected.
Holding:
Facts are not copyrightable; only creative selection/arrangement is
Minimal originality required
Relevance:
AI trained on musical datasets: using raw musical notes or chord sequences (as “facts”) may not infringe
Creative arrangements reproduced verbatim may still be infringement
3. Authors Guild v. Google, Inc. (2015, USA)
Authors Guild v. Google, Inc.
Facts:
Google scanned books for indexing and snippet display.
Holding:
Transformative use can qualify as fair use
Courts consider purpose, nature, amount, and market effect
Relevance:
AI music trained on copyrighted songs may qualify as transformative if outputs are new compositions and not exact copies
4. Bridgeman Art Library v. Corel Corp. (1999, USA)
Bridgeman Art Library v. Corel Corp.
Facts:
Photographic reproductions of public domain artworks claimed as copyrightable.
Holding:
Faithful reproductions of public domain works do not create new copyright
Only creative transformation is protected
Relevance:
AI music outputs based on public domain compositions (e.g., classical music) → copyright may apply only to new creative additions
5. Oracle America, Inc. v. Google, Inc. (2016, USA)
Oracle America, Inc. v. Google, Inc.
Facts:
Google used Java APIs to create Android.
Holding:
APIs are copyrightable, but transformative reuse may qualify as fair use
Relevance:
AI music tools implementing functional frameworks for music generation → lawful if independently coded and outputs are transformative
6. Thaler v. USPTO – DABUS AI Case (2021, USA/UK)
Thaler v. USPTO
Facts:
AI-generated inventions and creative works claimed to have AI authorship.
Holding:
US and UK courts: AI cannot be listed as author; human intervention required for copyright
Human selection, arrangement, or guidance establishes authorship
Relevance:
AI-assisted music composition: copyright may exist only if a human selects, edits, or directs the AI output
7. SAS Institute Inc. v. World Programming Ltd. (2013, EU)
SAS Institute Inc. v. World Programming Ltd.
Holding:
Functional aspects of software are not protected, only the expression of code
Independent reimplementation of functionality is lawful
Relevance:
Music generation AI: algorithms can be reimplemented independently without infringement
III. Key Legal Principles for AI Music Tools
Software Protection: Only the source code, not the algorithm itself, is protected.
Authorship Requirement: AI cannot hold copyright; humans guiding outputs can claim authorship.
Derivative Works: Outputs closely imitating copyrighted songs may infringe.
Training Data: Using copyrighted music to train AI may require license, unless fair use/transformative.
Original Outputs: Fully original compositions generated by AI under human guidance may be copyrighted.
IV. Practical Guidelines for AI Music Platforms
Use licensed or public domain music for training AI
Avoid reproducing large portions of copyrighted songs verbatim
Include human creative input in selection, arrangement, or curation
Protect software through copyright, but understand algorithms themselves are not protected
Monitor output for similarity to copyrighted works
V. Conclusion
AI-assisted musical composition sits at the intersection of:
Copyright in musical works
Copyright in software and algorithms
Derivative works doctrine
Authorship and moral rights
Key cases shaping the field:
Thaler v. USPTO (AI cannot be author; human intervention needed)
Computer Associates Int’l, Inc. v. Altai, Inc. (software expression vs idea)
Feist Publications v. Rural Telephone Service (facts vs expression)
Authors Guild v. Google, Inc. (transformative use and fair use)
SAS Institute Inc. v. World Programming Ltd. (functional software can be independently reimplemented)
RELATED Blog
comments