AI-Generated Content Ownership Claims in Europe

1. Introduction

AI-generated content ownership claims concern disputes over who, if anyone, legally owns copyright or other intellectual-property rights in material produced wholly or partly by an artificial-intelligence system.

The issue arises with:

AI-generated text;

AI-generated images and artwork;

AI-generated music;

AI-generated video;

AI-generated software code;

AI-generated designs;

AI-generated photographs;

AI-generated databases;

AI-generated translations;

AI-generated educational materials; and

commercially generated synthetic content.

The central European question is:

Can content generated by an AI system qualify for copyright, and if so, who owns the resulting rights?

There is currently no comprehensive EU rule granting copyright ownership merely because a person operates an AI system. EU copyright law traditionally links copyright protection to a human intellectual contribution. Consequently, the precise outcome depends upon the nature and extent of human creative intervention.

A crucial distinction must be made between:

ownership of the AI system;

ownership of the training data;

copyright in the AI output;

copyright in the human contributions to the output;

contractual rights concerning the output; and

rights in underlying or infringing material incorporated into the output.

2. European Legal Framework

The principal framework consists of:

EU copyright directives;

the Information Society Directive;

the Software Directive;

the Database Directive;

the Digital Single Market Directive;

national copyright laws;

EU Charter rights;

contractual law;

trade-secret law; and

increasingly, the EU AI Act.

The EU AI Act is important for AI governance and transparency, but it does not create a general rule saying that the user of generative AI automatically owns copyright in every output.

Copyright ownership remains principally governed by copyright law.

3. The Human-Creator Principle

The most important principle is that European copyright traditionally protects an original intellectual creation reflecting the author's own personality and free and creative choices.

Therefore, the difficult question is:

Who made the creative choices?

If an AI system independently determines:

composition;

wording;

colours;

structure;

melody;

arrangement; and

final expression,

there may be difficulty identifying a human author.

Conversely, if a human makes substantial creative decisions concerning the output, copyright may arise in the resulting human-created expression.

4. Case 1 — Infopaq International A/S v Danske Dagblades Forening

Court: Court of Justice of the European Union
Case: C-5/08
Year: 2009

Facts

Infopaq operated a media-monitoring service that scanned newspapers and extracted short sequences of words from articles.

The question was whether relatively small portions of text could fall within copyright protection.

Decision

The CJEU established an important EU copyright standard concerning an author's own intellectual creation.

Principle

Copyright protection can arise where the material represents the author's own intellectual creation.

Relevance to AI-generated content

This principle is foundational for AI-output disputes.

The question is not simply:

“Was the output produced by sophisticated technology?”

Instead, courts need to ask:

Does the protected expression embody a human author's own intellectual creation?

An AI-generated paragraph may therefore not automatically receive copyright merely because it is novel.

Conversely, human selection and arrangement surrounding AI-generated material may itself involve protectable creativity.

5. Case 2 — Painer v Standard VerlagsGmbH and Others

Court: CJEU
Case: C-145/10
Year: 2011

Facts

The dispute concerned photographs and whether portrait photography could satisfy the originality requirement.

Decision

The CJEU emphasized that a photograph may be protected where the photographer has been able to express creative abilities through making free and creative choices.

Principle

Creative choices concerning:

framing;

lighting;

composition;

setting;

timing; and

other expressive elements

can demonstrate an author's intellectual creation.

AI relevance

This case is extremely useful for AI-generated images.

Consider two scenarios.

Scenario A

A user enters:

“Generate a realistic photograph of a mountain.”

The AI independently determines essentially everything.

Human authorship may be difficult to establish.

Scenario B

A human:

designs the composition;

selects the subject;

generates multiple versions;

makes detailed creative selections;

edits lighting;

changes composition;

combines elements;

performs substantial post-processing.

The resulting human contribution may potentially satisfy the originality requirement.

6. Case 3 — Football Dataco Ltd v Yahoo! UK Ltd

Court: CJEU
Joined Cases: C-604/10
Year: 2012

Facts

The dispute concerned football fixture databases and whether the effort and skill involved in producing the database were sufficient for copyright protection.

Principle

Copyright protection depends on the presence of intellectual creativity rather than merely the expenditure of labour, money or effort.

AI relevance

This is particularly important because AI companies may argue:

“We spent enormous resources training and operating the model.”

That expenditure alone does not necessarily establish copyright in every generated output.

Similarly, a user spending hours crafting prompts does not automatically establish copyright.

The legally significant issue remains the creative intellectual contribution.

7. Case 4 — SAS Institute Inc. v World Programming Ltd

Court: CJEU
Case: C-406/10
Year: 2012

Facts

The dispute involved computer programs and the scope of copyright protection for software functionality, programming languages and program interfaces.

Principle

Copyright does not protect ideas and functionality as such in the same way that it protects original expression.

AI relevance

The distinction is important when claiming ownership over AI outputs.

A person generally cannot claim exclusive copyright merely because they developed:

a concept;

an idea;

a prompt;

a functional objective;

a style;

a mathematical method.

Copyright concerns protectable expression rather than every underlying idea or method.

For example:

“Write a detective story involving an unreliable narrator”

is an idea or instruction.

The resulting particular human-authored expressive elements may be protectable, but the general concept is not automatically owned.

8. Case 5 — Levola Hengelo BV v Smilde Foods BV

Court: CJEU
Case: C-310/17
Year: 2018

Facts

The dispute concerned whether the taste of a food product could qualify as a copyrighted work.

Decision

The CJEU emphasized the importance of a sufficiently identifiable and precise subject matter for copyright protection.

Principle

Copyright protection requires a work that can be identified with sufficient precision and objectivity.

AI relevance

This principle creates an important problem for generative AI.

Suppose an AI model produces an output that changes depending upon:

the prompt;

temperature;

model version;

random seed;

context;

external data; and

subsequent processing.

The claimant must still identify what precise work is supposedly protected.

A vague claim such as:

“I own the AI's style”

is therefore considerably weaker than:

“I own the specific original expressive elements that I personally created and incorporated into the final work.”

9. Case 6 — Cofemel — Sociedade de Vestuário SA v G-Star Raw CV

Court: CJEU
Case: C-683/17
Year: 2019

Facts

The case concerned the copyright status of clothing designs.

Principle

The CJEU reiterated the EU originality standard: a work must be an expression of the author's own intellectual creation.

AI relevance

AI-generated designs illustrate the problem clearly.

A fashion designer may use AI to generate hundreds of possible designs.

The fact that AI generated the alternatives does not necessarily mean that every generated design belongs to the designer.

However, the designer's subsequent:

selection;

modification;

arrangement;

combination;

alteration; and

artistic direction

may produce protectable human-created expression.

10. Case 7 — Brompton Bicycle Ltd v Chedech/Get2Get

Court: CJEU
Case: C-833/18
Year: 2020

Facts

The case concerned the copyright protection of the appearance of a functional bicycle design.

Principle

A design can potentially qualify for copyright even though it has a functional purpose, provided that it reflects original creative choices and is not dictated exclusively by technical considerations.

AI relevance

This is relevant to AI-generated:

product designs;

engineering designs;

architectural concepts;

industrial designs; and

functional objects.

The fact that an AI-generated design is technically useful does not automatically prevent copyright.

But the claimant still needs to establish the required human creative contribution.

11. Case 8 — Nintendo Co. Ltd v PC Box Srl

Court: CJEU
Case: C-355/12
Year: 2014

Facts

The case concerned copyright protection and technological protection measures surrounding video games.

Principle

EU copyright law protects qualifying creative works while balancing intellectual-property rights against legitimate interests and technological realities.

AI relevance

AI-generated videogame content raises similar questions involving:

AI-generated characters;

textures;

music;

dialogue;

environments;

code; and

game mechanics.

Ownership may have to be separated component by component.

12. Case 9 — SAS Institute: Software and AI Code

The SAS Institute judgment is particularly important for AI-generated computer code.

An AI coding assistant may generate:

function calculateTax(income) {    ... }

The user cannot necessarily claim copyright over every output simply because they entered the prompt.

The legal analysis should instead ask:

Did the human exercise creative choices?

Did the human select and arrange the code?

Is the resulting code original?

Does the output reproduce protected third-party code?

Is the claimed subject matter expression or merely functionality?

13. Case 10 — Infopaq and AI-Generated Text

The Infopaq principle is also relevant to generative writing.

Consider three levels:

Level 1 — Pure AI output

The user provides a simple prompt and accepts the result unchanged.

Copyright ownership may be uncertain or unavailable under the human-authorship requirement.

Level 2 — Human selection

The user generates 50 outputs and selects one.

Selection alone may or may not provide sufficient creative authorship depending upon the circumstances.

Level 3 — Substantial human rewriting

The user:

restructures the text;

rewrites paragraphs;

develops characters;

changes arguments;

creates original passages;

combines multiple outputs.

The final work may contain substantial human-authored expression.

14. The Most Important Ownership Distinction

AI disputes should distinguish:

A. Ownership of the AI model

Usually belongs to the developer or provider under applicable intellectual-property and contractual arrangements.

B. Ownership of training material

May belong to thousands or millions of third parties.

C. Ownership of the output

Depends upon applicable copyright law and human authorship.

D. Ownership of human modifications

May belong to the human creator if sufficiently original.

E. Contractual rights

May determine what the AI provider permits the user to do with outputs, even where copyright itself is uncertain.

15. Is the AI Itself an Author?

Under current European copyright principles, treating the AI system itself as the copyright author is highly problematic.

Copyright traditionally presupposes a human or legally recognized author.

Therefore:

AI-generated content is not automatically copyrighted merely because it was created by a sophisticated AI model.

The stronger legal question is:

Which human, if any, made the creative choices embodied in the work?

16. Prompting and Copyright

A difficult question is whether a sophisticated prompt creates authorship.

For example:

“Create a cinematic oil painting of a futuristic city at sunset, with three towers, a river in the foreground, a specific perspective, warm atmospheric lighting, and a particular symbolic composition.”

The prompt demonstrates creative intent.

But creative intent is not necessarily the same as creative expression.

The critical question is whether the human's creative choices are actually expressed in the final work.

A highly detailed prompt may strengthen an authorship argument, but it does not automatically establish copyright ownership.

17. Human Editing of AI Output

Human editing is one of the strongest potential routes to copyright protection.

Suppose AI generates a 1,000-word article.

The human then:

rewrites 600 words;

changes the structure;

adds original analysis;

removes AI errors;

creates new examples;

develops the argument;

selects the final organization.

The resulting work may contain protectable human-authored expression.

However, the copyright would not necessarily cover purely AI-generated portions merely because those portions were incorporated into a larger work.

18. Derivative and Composite Works

AI-generated works frequently contain a mixture of:

Human material + AI output + third-party material.

For example:

Human photograph + AI background + copyrighted character + human editing.

Ownership then becomes fragmented.

The claimant may own:

the original photograph;

the original modifications;

certain arrangement choices;

while lacking exclusive rights over:

third-party copyrighted elements;

purely AI-generated elements; or

public-domain material.

19. AI Output That Copies Existing Copyrighted Material

This is a separate issue from ownership.

Even if an individual has some rights in an AI-generated output, the output may infringe someone else's copyright.

For example:

User generates an image → AI reproduces substantial protected expression from an existing artwork.

The user's potential ownership claim does not automatically eliminate the original creator's rights.

This creates two separate questions:

Question 1

Does the user own copyright in the output?

Question 2

Does the output infringe somebody else's copyright?

Both questions must be independently answered.

20. AI Training and Ownership Claims

Training-data disputes must also be separated from output ownership.

An AI company may use copyrighted works during model development.

This can produce disputes concerning:

reproduction;

text and data mining;

licensing;

database rights;

contractual restrictions;

exceptions to copyright;

opt-out mechanisms; and

commercial exploitation.

But:

A dispute over whether training was lawful does not automatically determine who owns the final AI output.

They are legally distinct questions.

21. AI-Generated Music

Music creates particularly complex ownership problems.

AI may generate:

melody;

harmony;

rhythm;

lyrics;

instrumentation;

vocals.

Suppose a human writes the lyrics but AI generates the melody.

Potentially:

Human lyrics → human copyright

AI-generated melody → uncertain or potentially unprotected absent sufficient human authorship

Final recording → potentially separate rights

There may also be neighbouring rights relating to the sound recording and performer contributions.

22. AI-Generated Images

For images, the most useful analytical distinction is:

Pure generation

Simple prompt → AI output → no meaningful human modification.

Copyright claim may be difficult.

Creative direction

Extensive human selection and artistic manipulation.

Potential human copyright becomes stronger.

Post-production

AI output + Photoshop/other editing + human compositional decisions.

Copyright may potentially attach to the human-authored modifications and overall original arrangement.

The Painer and Cofemel principles are particularly useful here.

23. AI-Generated Software

AI coding tools create several separate legal questions.

Question 1

Who owns the AI-generated code?

Question 2

Does the code satisfy copyright originality requirements?

Question 3

Does the code reproduce third-party copyrighted code?

Question 4

What does the developer's contract with the AI provider say?

Question 5

Who is responsible if the AI-generated code violates an open-source license?

The SAS Institute jurisprudence is particularly relevant when distinguishing protectable expression from functionality and ideas.

24. Contractual Ownership

Even where copyright law does not automatically provide ownership, a contract may allocate rights between:

AI provider;

user;

employer;

customer;

freelancer;

university; or

enterprise.

For example, an AI service agreement might state:

“The customer receives specified rights to use outputs.”

That contractual permission does not necessarily mean:

“The customer owns copyright in the output.”

Ownership and contractual permission are different legal concepts.

25. Employer–Employee AI Content

Suppose an employee uses an AI tool to create:

advertising;

software;

illustrations;

reports;

product designs.

Ownership may depend upon:

national employee-copyright law;

employment contracts;

the employee's human creative contribution;

employer-assignment provisions;

the AI provider's terms.

The fact that the company paid for the AI subscription does not automatically answer the copyright question.

26. Evidence in AI Ownership Litigation

Courts may examine:

prompts;

prompt history;

model version;

generation logs;

successive outputs;

human editing records;

drafts;

metadata;

source files;

post-processing history;

contracts;

AI-provider terms;

training information where relevant;

evidence of third-party similarity.

Prompt history can become particularly important.

A claimant saying:

“I created this artwork”

may need to establish exactly what they contributed.

Evidence showing:

Prompt → 100 generated images → 5 selected → extensive editing → final image

may provide a much stronger authorship narrative than:

One prompt → one output → published unchanged.

27. Defenses Against AI Ownership Claims

A defendant may argue:

1. No human authorship

The output was generated autonomously.

2. Insufficient originality

The human contribution was too limited.

3. Idea rather than expression

The claimant is attempting to monopolize an idea, concept or style.

4. Third-party infringement

The output incorporates protected material belonging to another person.

5. Contractual restriction

The relevant contract allocates rights differently.

6. Public-domain material

The material claimed as exclusive is already free for use.

7. Lack of identifiable work

The claimant cannot precisely identify what protected expression they created.

28. Remedies

Potential remedies vary depending upon the dispute.

They can include:

declaration of ownership;

injunction;

damages;

account of profits;

removal of infringing material;

delivery or destruction of infringing copies;

correction of ownership records;

contractual remedies;

licensing arrangements.

For purely AI-generated material with uncertain copyright status, however, the most important dispute may be whether copyright exists at all.

29. The EU AI Act and Copyright Transparency

The EU AI Act is increasingly important for generative AI providers.

Providers of general-purpose AI systems have transparency and copyright-related obligations, including obligations concerning policies and information relating to training content.

However, the AI Act should not be misunderstood as creating a simple rule:

“The person who prompts an AI owns the output.”

It does not replace the EU copyright originality framework.

30. Practical Legal Test

A European court assessing an AI-generated content ownership dispute can be approached through the following questions:

Step 1

What exactly is being claimed as the copyrighted work?

Step 2

Was the work generated entirely by AI or partly by a human?

Step 3

What creative choices did the human make?

Step 4

Are those choices expressed in the final work?

Step 5

Is the human contribution sufficiently original?

Step 6

Can the protected expression be identified precisely?

Step 7

Does the output contain third-party protected material?

Step 8

What does the AI-provider contract provide?

Step 9

Are separate rights involved—for example, database, software, design or neighbouring rights?

Step 10

What remedy follows from the particular right that has been established?

31. Comparative Case Table

32. Six Most Important Principles

The case law collectively supports six major principles.

Principle 1 — Originality is central

Infopaq and Cofemel establish the importance of the author's own intellectual creation.

Principle 2 — Creative choices matter

Painer demonstrates that identifiable creative choices can establish originality.

Principle 3 — Effort alone is insufficient

Football Dataco shows that investment and labour do not automatically create copyright.

Principle 4 — Ideas are different from expression

SAS Institute is important for separating protectable expression from functionality and ideas.

Principle 5 — The work must be identifiable

Levola Hengelo emphasizes the need for a sufficiently identifiable subject matter.

Principle 6 — Functionality does not automatically destroy protection

Brompton Bicycle demonstrates that functional subject matter may still contain copyright-protected expression where genuine creative choices exist.

33. Core Problem: “AI Ownership” Is Often the Wrong Question

In practice, the better sequence is:

Was there a copyright-protected work?

↓

Who made the relevant creative contribution?

↓

Was that contribution sufficiently original?

↓

What part of the final output embodies that contribution?

↓

Does the output incorporate third-party rights?

↓

What contractual rights exist?

↓

What remedies follow?

This is considerably more accurate than simply asking:

“Who owns AI-generated content?”

34. Conclusion

European law currently does not provide a straightforward rule that every person who prompts an AI system automatically owns copyright in the resulting material.

The strongest European copyright authorities—particularly Infopaq, Painer, Football Dataco, SAS Institute, Levola Hengelo, Cofemel and Brompton Bicycle—point toward a human-centred originality analysis.

The decisive consideration is generally the existence of human intellectual and creative choices, rather than:

the sophistication of the AI;

the amount spent on the AI;

the number of prompts;

the computational effort;

the commercial value of the output; or

the fact that the user pressed the “generate” button.

Accordingly, a useful rule for European AI ownership disputes is:

Purely autonomous AI output presents the greatest difficulty for copyright ownership; substantial human creative selection, arrangement, modification or transformation creates a substantially stronger basis for copyright in the human-created elements.

Finally, copyright ownership and contractual permission must be kept separate. Even where copyright protection is uncertain, a contract may grant a user extensive rights to exploit an AI output; conversely, contractual permission to use an output does not necessarily establish that the user owns copyright in every element of it.