Copyright Implications For AI-Assisted Literary Translation Projects.
1️⃣ Is a Translation Protected by Copyright?
Core Principle
A translation is generally considered a derivative work because it is based on a pre-existing copyrighted text. Therefore:
If the original work is still under copyright → permission from the copyright holder is required.
If the original work is in the public domain → no permission is required, but the translation itself may receive separate copyright protection if sufficiently original.
📌 Key Case 1: Anderson v. Stallone (1989, U.S.)
Although not about translation specifically, this case is important for derivative works. The court held that unauthorized derivative works (a screenplay based on Rocky) cannot obtain copyright protection because they infringe the original work.
Relevance to AI Translation:
If AI translates a copyrighted novel without permission:
The translation is an infringing derivative work.
The translator (human or AI user) cannot claim valid copyright protection over it.
📌 Key Case 2: Salinger v. Random House (1987, U.S.)
This case dealt with unpublished letters by J.D. Salinger being quoted in a biography. The court emphasized that even transformation (like paraphrasing or rearranging) can infringe if it appropriates expressive content.
Relevance:
AI-generated translations reproduce expressive elements in another language. Even though the wording changes, the creative expression is preserved. Thus, translation is not a mere idea—it remains protected expression.
2️⃣ Does AI-Generated Translation Qualify for Copyright Protection?
Most jurisdictions require human authorship.
📌 Key Case 3: Burrow-Giles Lithographic Co. v. Sarony (1884, U.S.)
The U.S. Supreme Court held that copyright protects works originating from “the creative powers of the mind.”
Relevance:
Copyright requires human intellectual creation. If a translation is produced autonomously by AI with no meaningful human input, it may fail the authorship requirement.
📌 Key Case 4: Naruto v. Slater (2018, U.S.)
Known as the “Monkey Selfie” case. The Ninth Circuit held that non-humans cannot own copyright.
Relevance:
AI systems cannot be authors. If AI independently produces a translation:
The AI cannot own copyright.
The output may enter the public domain unless a human exercises sufficient creative control.
📌 Key Case 5: Thaler v. Perlmutter (2023, U.S.)
A U.S. court confirmed that works generated solely by AI without human authorship are not copyrightable.
Relevance to Literary Translation:
If a user simply inputs a text and the AI fully translates it with no creative choices by a human:
The translation likely lacks copyright protection in the U.S.
However, if a human edits, restructures, or creatively intervenes, copyright may attach to the human contribution.
3️⃣ Originality in Translation
Translation involves interpretive choices: tone, nuance, cultural adaptation, idiomatic shifts.
📌 Key Case 6: Feist Publications v. Rural Telephone Service (1991, U.S.)
The Supreme Court clarified that originality requires:
Independent creation
A minimal degree of creativity
Relevance:
A human-guided translation generally satisfies originality.
But a purely mechanical or automated translation may fail if it lacks independent creative input.
📌 Key Case 7: Walter v. Lane (1900, UK)
The House of Lords held that verbatim reporters’ transcripts of speeches were original works due to skill and labor.
Relevance:
Historically, courts recognized “skill and labor.”
However, modern EU law requires intellectual creation, not mere labor. So simple mechanical AI translation without intellectual choice may not qualify.
4️⃣ EU Standard: “Author’s Own Intellectual Creation”
📌 Key Case 8: Infopaq International A/S v. Danske Dagblades Forening (2009, CJEU)
Infopaq International A/S v. Danske Dagblades Forening
The Court of Justice of the European Union held that copyright subsists where the work reflects the author’s “own intellectual creation.”
Relevance:
For AI-assisted translation in the EU:
If a human meaningfully shapes the translation → protected.
If AI acts autonomously → unlikely to qualify as “intellectual creation.”
5️⃣ Moral Rights in Literary Translation
In civil law countries (e.g., France, Germany):
Authors retain moral rights (right of attribution, integrity).
Even authorized translations must not distort the original.
📌 Key Case 9: Snow v. The Eaton Centre Ltd. (1982, Canada)
Snow v. The Eaton Centre Ltd.
An artist successfully argued that modifications to his sculpture violated his moral rights.
Relevance to AI Translation:
If AI alters tone, removes political nuance, or distorts meaning:
The original author could claim violation of moral rights.
This is especially important in jurisdictions recognizing strong moral rights.
6️⃣ Fair Use and Machine Training
AI systems are trained on copyrighted works.
📌 Key Case 10: Authors Guild v. Google, Inc. (2015, U.S.)
Authors Guild v. Google, Inc.
The Second Circuit held that Google’s book scanning for search functionality was fair use because it was transformative and did not substitute for the books.
Relevance:
AI training on books may be considered transformative (depending on jurisdiction). However:
Generating full translated texts may compete with the original market.
That could weigh against fair use.
7️⃣ Substantial Similarity and AI Outputs
📌 Key Case 11: Nichols v. Universal Pictures (1930, U.S.)
Nichols v. Universal Pictures
Judge Learned Hand developed the abstraction test for determining substantial similarity.
Relevance:
In translation:
The plot, characters, and expressive structure are preserved.
Therefore, translations are almost always substantially similar to the original.
Unauthorized AI translations are legally risky.
8️⃣ Liability in AI-Assisted Translation
Potentially liable parties:
The user who uploads copyrighted text
The publisher distributing the translation
Possibly the AI platform (depending on safe harbor protections)
Relevant doctrines:
Secondary liability
Contributory infringement
Vicarious liability
The legal framework is still evolving.
9️⃣ Key Legal Questions in AI-Assisted Translation
A. Who Is the Author?
AI alone → likely no copyright.
Human edits + creative intervention → human author.
B. Is Permission Required?
If original is copyrighted → YES.
AI involvement does not eliminate infringement.
C. Who Owns the Translation?
If sufficiently human-authored → the human.
If purely AI-generated → possibly no one.
D. Are Moral Rights Implicated?
Yes, especially in EU and civil-law countries.
🔟 Practical Risk Scenarios
| Scenario | Legal Risk |
|---|---|
| Translating a public domain novel using AI | Low |
| Translating a copyrighted novel without permission | High |
| AI-assisted translation with heavy human editing | Moderate–Low (if authorized) |
| Publishing AI translation commercially | High risk without license |
Conclusion
AI-assisted literary translation sits at the intersection of:
Derivative work doctrine
Human authorship requirement
Originality standards
Moral rights protections
Fair use/fair dealing analysis
The dominant principle across jurisdictions remains:
Copyright protects human intellectual creativity. AI is a tool—not an author.
Therefore:
Permission is required for translating copyrighted works.
AI output without meaningful human input may lack protection.
Moral rights and substantial similarity doctrines remain central.
Liability risks increase with commercial distribution.
RELATED Blog
comments