Why AI Output Is Not Fair Use: Understanding Transformative Use in International Copyright Law

 

1. Introduction: “AI Output Looks Different — So Isn’t It Fair Use?”

One of the most common arguments defending AI-generated art is:

“The output looks different, so it must be fair use.”

However:

❌ Fair use is not about how different a work looks.

✔ Fair use evaluates purpose, character, transformation, and market impact.

And in almost every legal analysis so far:

AI output does not satisfy the fair use doctrine.

2. What Is Fair Use and the Transformative Use Test?

In U.S. law (the system AI developers rely on most), fair use is evaluated using four factors:

  1. Purpose and character of the use

  2. Nature of the copyrighted work

  3. Amount and substantiality used

  4. Effect on the market for the original

The most important factor is the first one:

Transformative Use

A use is transformative if:

  • it adds new meaning,

  • introduces new expression,

  • or repurposes the work for a different creative objective,

  • through human creativity.

3. Why AI Output Fails the Transformative Use Test

AI:

❌ has no intent

❌ has no creativity

❌ does not reinterpret meaning

❌ merely recombines statistical patterns from training data

It does not create new meaning or artistic commentary.

Transformative use in legal terms requires:

✔ human creativity

✔ intentional reinterpretation

✔ expressive transformation

AI output is therefore not transformative — it is derivative.

4. Real-World Cases Showing AI Output Is Not Fair Use

A. Getty Images vs. Stability AI

  • AI output resembled Getty-owned images

  • Some outputs even contained distorted Getty watermarks

  • Court concluded this was not transformation

  • But a form of unauthorized reproduction

B. Sarah Andersen vs. Midjourney & DeviantArt

  • AI generated comic-like images similar to Andersen’s style

  • No human creative reinterpretation

  • The output competed directly with the original market

  • Not transformative

  • Not fair use

5. Why AI Output Fails All Four Fair Use Factors

A. Purpose and Character — Fails

AI-generated works are:

  • used commercially

  • designed to replace human artists

  • not created through human interpretation

No meaningful transformation occurs.

B. Nature of the Work — Fails

AI uses:

  • highly creative works

  • illustrations, photography, fine art

  • works that receive the strongest copyright protection

C. Amount Taken — Fails

AI systems do not extract “a small portion.”

They ingest:

→ the entire work

→ including structure, patterns, stylistic identity

This level of copying is far beyond what fair use allows.

D. Market Harm — Fails

AI-generated works compete directly with:

  • illustrators

  • concept artists

  • photographers

  • designers

AI can produce thousands of works in seconds, severely harming the original artist’s market.

Market harm alone can defeat a fair use claim.

6. The European Union: Even Stricter Than the U.S.

The EU does not use “fair use.”
It uses narrowly defined Exceptions and Limitations, and:

❌ There is no exception allowing style imitation

❌ There is no exception allowing derivative works via AI

❌ Commercial AI training without permission violates EU law

The EU AI Act also requires:

  • dataset transparency

  • copyright-respecting training

  • documentation obligations

Under European law, AI outputs resembling copyrighted works are almost certainly infringing.

7. Indonesia: AI Output Cannot Be Fair Use

Indonesia does not have the fair use doctrine.

Instead, Indonesia uses fair dealing, which is far narrower and limited to:

✔ education
✔ research
✔ library use
✔ citation
✔ non-commercial purposes

AI output, however:

  • is commercial

  • uses entire works

  • replaces human creators

  • harms the artist’s economic rights

  • originates from training on unlicensed data

Therefore:

AI output cannot qualify as “Penggunaan Wajar” (fair dealing) in Indonesia.

8. Conclusion: AI Output Is Not Fair Use

❌ AI is not creative → fails the transformative test

❌ AI harms the market → fails factor 4

❌ AI uses full works → fails factor 3

❌ AI uses highly protected works → fails factor 2

❌ AI output replaces human labor → commercial exploitation

❌ AI output often resembles original works → derivative, not transformative

Therefore:

AI output cannot be justified under fair use.

AI output requires permission, licensing, and attribution.

If AI is to operate legally and ethically, developers must ensure:

✔ licensed datasets
✔ consent-based training
✔ artist compensation
✔ output transparency

Comments

Post a Comment

Popular posts from this blog

Use of Stock Images, Icons, and UI Assets in Games: Legal Rules Developers Must Know

 Game developers frequently use: icon packs stock photos illustration bundles UI kits button packs overlay graphics mockup assets template HUDs These assets speed up development — but they are also among the top causes of copyright violations, DMCA takedowns, and publisher rejections . Why? Because many developers do not read the licenses . In reality: Visual assets are NOT free to use just because they are easy to download. This article explains the legal rules and risks around stock images, icons, and UI kits in game development. ⭐ 1. Stock Images and Icons Are Copyrighted Works All visual assets — including icons, illustrations, UI elements, and stock photos — are protected by copyright. Just because something is: downloadable, free, shared publicly, available on Google Images does NOT mean it is legal to use. Any visual asset requires proper licensing for: commercial use, embedding in a game, redistribution to playe...
Read more

Music Copyright in Games: Licensing, Usage Rules, and Legal Risks for Developers

 Music is one of the most powerful elements in game design: background music (BGM) battle themes ambience UI sound sound effects (SFX) voiceovers However — Music is also one of the MOST legally sensitive assets in game development. Many developers underestimate how strict music copyright law is. A single mistake can cause: takedowns publisher rejection console certification failure delayed global release DMCA strikes expensive legal disputes This article explains how music licensing really works and why studios must handle it carefully. ⭐ 1. You Cannot Use Music from YouTube, Spotify, or Streaming Platforms One of the most common misunderstandings: “If it’s online, I can use it.” Absolutely NOT. Using music from: YouTube Spotify Apple Music SoundCloud TikTok is 100% illegal for game use. This violates: composition copyright master rights mechanical rights synchronization rights Games always req...
Read more

What Makes AI Training Data Illegal? A Breakdown of the Most Common Dataset Violations in AI Development

  1. Introduction: The Real Legal Problem in AI Is Not the Output — It’s the Dataset Most legal issues in AI do not originate from: the AI-generated output, style imitation, or model errors. The biggest legal risk begins before the model even produces anything — during dataset collection and training . If the dataset is illegal, it can make the entire AI model legally compromised : ❌ copyright infringement ❌ violation of moral rights ❌ privacy violations ❌ breach of publicity rights ❌ breach of platform Terms of Service ❌ misuse of licensed or contract-protected content ❌ inclusion of sensitive or illegal data And because training is performed by the developer , the developer — not the AI or the user — is legally responsible . 2. What Makes an AI Dataset Illegal? A dataset becomes illegal when one or more of the following conditions apply: A. Contains Copyrighted Works Without Permission This is the most common and widely litigated violation. If a...
Read more