Copyright Implications In AI-Based Historical Restoration.
1. Background: AI-Based Historical Restoration
AI-based historical restoration involves using AI to restore, reconstruct, or enhance old works, such as:
Old films and photographs (colorization, frame interpolation)
Historical audio recordings (noise reduction, voice enhancement)
Artworks and manuscripts (reconstructing damaged or faded areas)
While transformative and culturally valuable, this raises copyright and moral rights questions, especially when:
The original works are still under copyright.
AI generates outputs that resemble the original work too closely.
Restoration involves creative interpretation.
Key legal issues include authorship, derivative works, fair use, and moral rights.
2. Key Legal Issues
Derivative Works: Restoration may create derivative works. Unauthorized derivatives of copyrighted materials can infringe.
Authorship and Originality: If AI generates creative elements (color, textures, reconstructed scenes), who owns the copyright? The programmer, the restorer, or nobody?
Moral Rights: In some jurisdictions (like Europe), authors retain the right to object to modifications that distort their work.
Fair Use / Public Interest: Restoration may be defended as transformative, especially for historical or educational purposes.
3. Case Law Illustrations
Case 1: Bridgeman Art Library v. Corel Corp. (1999, SDNY)
Issue: Corel reproduced exact photographic images of public domain artworks.
Ruling: Exact photographic copies of public domain works are not copyrightable because they lack originality.
Relevance: In historical restoration, if AI reproduces a work exactly, the output may not have copyright, but creating new creative enhancements may. AI restorations that are purely technical (like cleaning noise from an image) may not generate copyrightable content.
Case 2: Authors Guild v. Google, Inc. (2015, SDNY)
Issue: Google scanned books and created searchable text.
Ruling: Use was fair use because it was transformative (search function) and did not replace the original.
Relevance: AI-based restoration may qualify as fair use if it adds significant value (enhancement, accessibility) without substituting for the original work.
Case 3: Warhol Foundation v. Goldsmith (2021, SDNY)
Issue: Warhol created silkscreens based on Goldsmith’s photograph.
Ruling: Warhol’s work was not transformative enough—substantial copying remained.
Relevance: If AI reconstruction reproduces copyrighted images too closely (e.g., reconstructing a historical film frame using copyrighted footage), it may infringe.
Case 4: Feist Publications v. Rural Telephone Service (1991, US Supreme Court)
Issue: Copying factual content from a telephone directory.
Ruling: Only original arrangement or selection can be copyrighted.
Relevance: AI-based restoration that organizes or reconstructs historical works can be protected only if creative decisions are made, not just mechanical restoration.
Case 5: Naruto v. Slater (Monkey Selfie, 2018, 9th Circuit)
Issue: Can non-humans hold copyright?
Ruling: Only humans can own copyright.
Relevance: AI cannot claim authorship, so ownership defaults to the human operator who directed or supervised restoration.
Case 6: Cariou v. Prince (2009, SDNY)
Issue: Richard Prince used Patrick Cariou’s photographs in artworks.
Ruling: Some of Prince’s uses were transformative, protected under fair use.
Relevance: AI restoration that adds creative interpretation (like colorization or stylistic reconstruction) may be considered transformative.
Case 7: Kienitz v. Sconnie Nation LLC (2014, 7th Circuit)
Issue: Parody and photo modification.
Ruling: Even small modifications can be fair use if sufficiently transformative.
Relevance: Minor AI alterations of historical works (noise reduction, missing frame reconstruction) may be fair use if they transform or reinterpret the original.
4. Practical Implications for AI-Based Historical Restoration
Rights Clearance: Check whether the original work is still under copyright.
Human Oversight: Document human creative choices in AI restoration to claim authorship.
Transformative Output: Ensure restoration adds value or changes meaning, not just reproduces the original.
Moral Rights Consideration: Particularly relevant in Europe, where modifications cannot harm the original author’s reputation.
Transparency: Clearly disclose AI’s role in restoration to avoid misattribution.
5. Conclusion
AI-based historical restoration lies in a nuanced legal landscape:
Purely technical restorations (exact replicas) may not be copyrightable (Bridgeman Art Library v. Corel).
Transformative enhancements, reconstructions, or interpretations may be protected if human authorship and creativity are involved (Cariou v. Prince, Feist v. Rural).
Unauthorized copying of copyrighted historical works risks infringement (Warhol Foundation v. Goldsmith).
AI cannot hold copyright, so human operators must ensure proper rights management (Naruto v. Slater analogy).
These cases collectively demonstrate that AI-based historical restoration requires careful attention to originality, human authorship, and transformative use to navigate copyright law safely.
RELATED Blog
comments