by lawyerfbc

on August 24, 2026

Generative artificial intelligence has fundamentally changed the way businesses create content.

A startup can now use artificial intelligence to generate:

articles,
advertising copy,
software code,
photographs,
illustrations,
logos,
product descriptions,
music,
videos,
presentations,
website designs,
social media posts,
contracts,
reports,
and other commercial materials

within seconds.

This technological development creates an increasingly important legal question:

Who owns the copyright in content generated by artificial intelligence?

Does the person who writes the prompt own the output?

Does the company paying for the AI subscription own it?

Does the developer of the artificial intelligence model own it?

Can the artificial intelligence itself be considered the author?

What happens when an employee creates AI-generated content while working for a startup?

And perhaps most importantly:

Can purely AI-generated content receive copyright protection at all under Turkish law?

These questions are particularly significant for technology startups because AI-generated materials are increasingly used as commercially valuable assets.

A startup may create its entire marketing campaign through generative AI. A software company may use AI-generated code inside its product. A gaming startup may produce characters and environments through image-generation systems. An advertising startup may create thousands of product images automatically.

If those materials are not legally protected in the way the founders expect, competitors may potentially copy them more easily.

At the same time, using AI-generated content can create a second and completely different risk:

The output may reproduce or infringe somebody else’s copyright.

Accordingly, AI copyright analysis involves two separate questions:

Can the AI-generated output itself be protected by copyright?
Does using the AI-generated output infringe another person’s existing rights?

These questions must not be confused.

Under Turkish law, copyright protection is principally governed by Law No. 5846 on Intellectual and Artistic Works, commonly referred to as the FSEK.

As of August 2026, the FSEK does not contain a specific rule declaring an artificial intelligence system to be an author or automatically assigning ownership of purely AI-generated material to the user or AI provider.

The existing framework therefore needs to be applied by examining concepts such as:

work,
author,
originality,
human creative contribution,
adaptation,
compilation,
employee-created works,
contractual transfers,
and infringement.

A legislative proposal submitted to the Turkish Parliament on 8 April 2026 specifically aims to update Law No. 5846 in response to copyright needs arising in the artificial intelligence era. However, as of 20 August 2026, that proposal remains before the relevant parliamentary commissions and has not become law.

This article examines who owns the copyright in AI-generated content under Turkish law, how human contribution affects copyright protection and what startups should consider before commercially using generative AI outputs.

What Is a Copyright-Protected Work Under Turkish Law?

The starting point is the definition of a protected work.

Under the FSEK, a work must carry the characteristics of its author and fall within one of the statutory categories of works.

The Ministry of Culture and Tourism explains that, for an intellectual and artistic product to receive copyright protection, it should generally:

result from intellectual effort,
carry the author’s individual characteristics,
have taken a concrete form,
and fall within one of the categories recognised by the statute.

The Ministry also states the fundamental rule that the author of a work is the person who creates it.

Article 1/B similarly defines a work as an intellectual and artistic product carrying the characteristics of its author and defines the author as the person who creates the work.

This structure becomes extremely important when generative AI is involved.

Can Artificial Intelligence Be the “Author” Under Turkish Copyright Law?

Under the current structure of the FSEK, there is a strong legal basis for concluding that an AI system itself cannot simply be treated as the author in the same way as a human creator.

Article 8 states:

“The author of a work is the person who creates it.”

The wider statutory framework is built around concepts such as:

the author’s personality,
moral rights,
attribution,
disclosure,
objection to modifications,
death of the author,
heirs,
and contractual exercise of economic rights.

These concepts are designed around human creators.

The FSEK also distinguishes the author from legal entities that may exercise economic rights in particular circumstances. For example, Article 18 allows employers, under the statutory conditions, to exercise rights over works created by employees while carrying out their duties.

This distinction reinforces an important point:

The company may exercise economic rights without necessarily being the natural creator of the work.

An artificial intelligence system has no legal personality, personal creativity in the legal sense, moral rights or capacity to enter into copyright transactions.

Therefore, simply stating that “Chatbot X is the author” would not fit comfortably within the existing Turkish copyright structure.

Does This Mean All AI-Generated Content Is Unprotected?

No.

This is where the analysis becomes more complex.

There is a significant difference between:

content generated entirely autonomously by AI

and

content created by a human with AI assistance.

AI can operate in different roles.

It can function as:

a technical tool,
an editing assistant,
an idea generator,
a drafting assistant,
or the predominant source of expressive content.

Scenario One: Pure AI Generation With Minimal Human Contribution

Consider the following prompt:

“Create a photograph of a futuristic city at sunset.”

The user clicks “Generate”.

The AI independently determines:

composition,
buildings,
lighting,
colours,
perspective,
details,
textures,
and final visual expression.

The user did not select those particular expressive elements.

The user merely communicated a general idea.

Under the traditional concepts of Turkish copyright law, there is a serious question whether the resulting image can be regarded as carrying the individual characteristics of a human author.

The FSEK’s definition of a work requires the product to carry the characteristics of its author, while Article 8 identifies the creator as the author.

Where the expressive choices are predominantly determined by the AI system rather than a human creator, claiming full copyright protection in the output may therefore be legally difficult under the current statutory structure.

This should not be confused with saying the image is illegal.

The image may be perfectly lawful to use.

The issue is whether the person using it can claim exclusive copyright ownership preventing everybody else from copying it.

Scenario Two: AI Used Merely as a Creative Tool

Now consider a different example.

A professional graphic designer creates an original composition.

The designer:

draws the basic layout,
determines the characters,
selects the colours,
creates the background,
defines perspective,
generates certain components through AI,
manually edits those components,
combines them with original drawings,
removes unwanted elements,
creates additional elements,
and makes significant creative decisions concerning the final composition.

In this situation, artificial intelligence is closer to a tool.

The final product may contain substantial human creative expression.

The relevant question becomes:

Can the human-created aspects satisfy the FSEK requirements independently of the AI-generated components?

Potentially, yes.

A court examining such a dispute would likely need to distinguish the human creative contributions from purely machine-generated material.

The fact that AI was used during the creative process should not automatically eliminate protection for the human-authored elements.

The Important Distinction: AI-Generated vs AI-Assisted

Startups should therefore distinguish:

AI-Generated Content

The AI system determines most or all expressive elements.

Human participation is minimal.

AI-Assisted Content

A human author makes sufficiently meaningful creative choices and uses AI merely as part of the creative process.

The second category is considerably easier to analyse under traditional copyright principles.

The legal focus should not be:

“Was artificial intelligence used?”

Instead, it should be:

“Which expressive elements were actually created by the human?”

Does Writing a Prompt Make the User the Author?

Not necessarily.

This is one of the most debated questions in generative AI copyright law.

Suppose a person spends ten minutes writing a detailed prompt describing:

visual style,
camera angle,
composition,
characters,
lighting,
colours,
environment,
and atmosphere.

Does that make the user the author of the resulting image?

Under Turkish law, the answer cannot currently be stated categorically for every case.

The prompt itself may contain creative expression.

But copyright in the prompt and copyright in the output are separate questions.

A person may own copyright in an original written prompt while not necessarily owning copyright in every expressive element produced unpredictably by the AI.

This distinction becomes clearer when one considers the difference between:

giving instructions

and

creating the resulting expression.

For example, a film director giving extensive instructions may still make legally relevant creative contributions because the director exercises control over the final audiovisual work.

By contrast, if a generative AI system independently determines major expressive details after receiving a general prompt, the causal connection between the prompt and the final expression may be weaker.

Comparative Perspective: The U.S. Copyright Office and Prompts

Although U.S. law does not determine Turkish copyright law, international developments are useful for understanding the debate.

In January 2025, the U.S. Copyright Office published Part 2 of its Artificial Intelligence Report and concluded that generative AI output may receive protection where sufficient human-authored expressive elements exist.

It also concluded that merely providing prompts to current generative AI systems does not ordinarily give the user sufficient control over the expressive elements produced by the AI.

The Office nevertheless emphasised that AI-assisted works may remain copyrightable where humans create, select, arrange or modify expressive elements.

This comparative approach is broadly consistent with the central question arising under the Turkish FSEK:

Where is the human author’s individual creative contribution?

Can a Very Detailed Prompt Itself Be Copyrighted?

Potentially.

A prompt is text.

If it contains sufficiently original expression and satisfies the statutory requirements for a work, it may theoretically receive protection as a literary work.

For example, compare:

“Create a red car.”

with a several-page original description containing:

characters,
dialogue,
narrative structure,
setting,
artistic choices,
and detailed original literary expression.

The second text has a much stronger possibility of independently qualifying for copyright protection.

However, ownership of the prompt does not necessarily mean ownership of every output generated from that prompt.

The prompt and output remain legally distinguishable.

What If the User Generates 100 Images and Selects One?

Selection alone does not necessarily transform the selected image into a human-authored work.

However, selection and arrangement can become legally important in more complex projects.

The FSEK expressly recognises compilation works where the selection and arrangement result from intellectual creativity, while preserving the rights relating to underlying works.

For example, a designer may:

generate hundreds of images,
select twenty,
arrange them into a structured visual story,
determine sequence,
add original text,
edit the images,
and create an overall artistic composition.

Even if individual AI-generated images have uncertain copyright status, the human-created selection and arrangement may potentially create separately protectable elements.

The extent of protection will depend on the actual creative contribution.

What If a Human Heavily Edits AI-Generated Content?

This can substantially strengthen the copyright argument.

Suppose AI generates a basic illustration.

A human artist then:

redraws facial features,
changes composition,
replaces the background,
adds original characters,
changes colours,
creates textures,
modifies lighting,
adds new artistic elements,
and integrates the result into a larger original work.

The final output may carry human creative characteristics that did not exist in the raw AI generation.

Copyright protection may therefore attach to those human-authored modifications or to the resulting work to the extent statutory requirements are satisfied.

Again, the analysis is not binary.

The question is not merely:

“AI or human?”

It is:

“Which parts reflect protected human creative expression?”

Who Owns AI-Generated Content Created by an Employee?

This is highly relevant to startups.

Suppose a startup employs a marketing designer.

The employee uses generative AI tools to prepare advertising visuals as part of ordinary employment duties.

The first question should be:

Is there a copyright-protected human work in the final material?

If the content is purely machine-generated and contains no sufficient human-authored expression, there may be no conventional copyright for either the employee or employer to own.

However, if the employee contributes sufficiently original human expression, Article 18 becomes relevant.

Article 18 provides that, unless otherwise understood from a special agreement or the nature of the work, rights over works created by employees while carrying out their duties are exercised by the employer.

Therefore, the legal sequence should be:

determine whether a protected work exists;
identify the human author;
determine whether it was created during employment duties;
analyse the employer’s rights under Article 18 and the employment agreement.

An employment agreement cannot magically create copyright in something that does not qualify for copyright protection in the first place.

What If a Freelancer Creates Content Using AI?

The same issue arises when startups hire freelancers.

Suppose a startup pays a freelancer to create fifty illustrations.

The freelancer generates all fifty through AI with minimal human modification.

The agreement says:

“The freelancer transfers all copyrights to the startup.”

There may be a fundamental problem.

A person cannot transfer a statutory copyright that does not exist.

Article 52 of the FSEK requires contracts concerning economic rights to be in writing and requires the relevant rights to be specified individually. Article 54 also provides important consequences where rights are acquired from a person who does not have authority to transfer them.

Therefore, startups should not assume that a broad IP assignment clause solves every AI copyright issue.

The company should ask:

Was there sufficient human authorship?
Who made the creative contribution?
Which rights actually exist?
Which rights are being transferred?
Does the freelancer have authority to transfer them?

AI Platform Terms Do Not Determine Copyright Law

Another common misconception is:

“The AI company’s Terms of Service say that users own their outputs, therefore copyright definitely belongs to us.”

Contractual terms are important.

They determine the legal relationship between:

the AI provider,
the user,
and sometimes the user’s company.

However, contractual language cannot necessarily create statutory copyright protection where the applicable copyright law does not recognise the output as a protected work.

For example, an AI provider may contractually say that it assigns whatever rights it may have in output to the user.

That can be commercially useful.

But the critical phrase is often effectively:

“whatever rights exist.”

The contract cannot guarantee that a court will recognise human authorship in every generated image, text or piece of music.

Startups should therefore analyse both:

the AI platform’s contractual terms; and
applicable copyright legislation.

Does Paying for a Premium AI Subscription Give Copyright Ownership?

No automatic rule says:

“The person who pays for the AI tool owns the copyright.”

Payment determines access to the service.

Similarly:

paying for Photoshop does not make Adobe the author of every image;
owning a camera does not determine authorship of every photograph;
paying for an AI subscription does not automatically determine authorship of every output.

The relevant issue remains the nature of the creative contribution.

What If the Startup Develops Its Own AI Model?

A startup developing its own model must distinguish several intellectual property layers.

These may include:

The AI Software

Source code may receive copyright protection.

Model Architecture and Technical Materials

Different intellectual property and confidentiality protections may apply.

Training Data

Individual training materials may contain third-party copyright.

Model Weights

Their precise legal treatment can involve complicated questions depending on the circumstances.

Prompts

Some prompts may contain protectable human expression.

Outputs

A startup should not treat the entire AI ecosystem as one single intellectual property asset.

AI Output Ownership and AI Training Are Separate Copyright Questions

This distinction is fundamental.

Suppose a startup asks an AI system to generate an illustration.

Question One:

Does the user own copyright in the illustration?

Question Two:

Was the model legally trained using copyrighted images?

These are different questions.

Even if the output itself does not receive copyright protection, the model provider may still face separate issues concerning training materials.

Likewise, even if the training process was lawful, that does not automatically mean the resulting output belongs to the user as copyright.

The two analyses should remain separate.

Can Publicly Available Online Content Be Freely Used to Train AI?

Startups should not assume:

“If it is on the internet, it is free to use for AI training.”

Online accessibility does not automatically eliminate copyright.

The legality of reproducing protected material for AI training requires analysis under the relevant copyright system, licences, statutory limitations and factual circumstances.

This issue is especially important for startups building their own models rather than merely using an external API.

In the European Union, for example, Article 4 of Directive 2019/790 provides a text-and-data-mining exception for lawfully accessible materials subject to important conditions, including the right holder’s ability to reserve relevant rights in an appropriate manner.

This EU rule should not simply be assumed to apply identically under Turkish copyright law.

A Turkish AI startup training its own model should conduct a separate dataset-rights analysis.

The EU AI Act Now Contains Copyright-Related Obligations for GPAI Providers

Turkish startups placing general-purpose AI models on the European Union market should also consider the EU Artificial Intelligence Act.

Article 53 requires providers of general-purpose AI models to establish a policy to comply with EU copyright law, including identifying and respecting reservations of rights made under Article 4(3) of the Digital Single Market Copyright Directive.

Providers must also prepare and publicly make available a sufficiently detailed summary of the content used for training according to the AI Office template.

The EU General-Purpose AI Code of Practice includes a dedicated Copyright Chapter designed to assist model providers in demonstrating compliance with these obligations.

These requirements concern training and model-provider compliance.

They do not establish a general rule stating that the person who writes a prompt automatically owns copyright in the resulting output.

AI-Generated Content Transparency Is Also Becoming Important

From 2 August 2026, Article 50 transparency obligations under the EU AI Act apply to certain AI-generated and manipulated content.

Among other things, providers of systems generating synthetic audio, image, video or text must satisfy certain machine-readable marking obligations, while deployers face disclosure requirements concerning deepfakes and specified AI-generated public-interest text.

Again:

Labelling content as AI-generated does not determine who owns copyright in it.

Transparency and ownership are separate questions.

Can AI-Generated Content Infringe Existing Copyright?

Yes.

This is perhaps the most important commercial risk.

The fact that an AI system generated an output does not guarantee that the output is legally safe.

For example, the output might reproduce:

a protected illustration,
substantial literary text,
source code,
a character,
a musical composition,
or other protected expression.

Suppose a user asks:

“Create an image identical to Artist X’s protected illustration.”

If the resulting image reproduces protected expression, a copyright dispute may arise.

The person commercially using the output cannot necessarily defend the claim simply by saying:

“The AI created it.”

Similarity Does Not Automatically Mean Copyright Infringement

At the same time, not every similarity constitutes infringement.

ideas,
concepts,
themes,
methods,
or general artistic concepts.

For example, producing:

“a detective story set in Istanbul”

does not necessarily infringe another detective novel merely because the general concept is similar.

The legal analysis should examine whether protected original expression has been reproduced.

This becomes difficult with generative AI because outputs may contain combinations of patterns learned from extremely large datasets.

Human legal review may therefore be necessary for commercially important outputs.

What About Creating Content “In the Style of” an Artist?

Requests such as:

“Generate a painting in the style of Artist X”

create complicated intellectual property and personality-right questions.

An artistic style as an abstract concept is not necessarily protected in the same way as a particular copyrighted work.

However, the resulting output may create problems if it reproduces specific protected elements of existing works.

Other legal issues may also arise concerning:

trademarks,
false endorsement,
personality rights,
unfair competition,
and misleading commercial conduct,

depending on how the content is used.

Startups should therefore avoid reducing the analysis to:

“Style is not copyrighted, so everything is safe.”

The actual output and commercial context should be examined.

AI-Generated Logos Create Additional Risks

AI-generated logos are increasingly popular among early-stage startups.

However, founders should be cautious.

Three separate issues arise.

First: Copyright

A purely AI-generated logo may have uncertain copyright status.

Second: Trademark

The logo may still potentially function as a trademark if it satisfies trademark-law requirements and does not conflict with earlier rights.

Trademark and copyright are separate legal systems.

Third: Third-Party Similarity

The AI may generate a logo resembling an existing trademark.

Therefore, a trademark clearance search should still be conducted before commercial launch.

The phrase:

“AI created the logo”

is not a substitute for trademark clearance.

Can an AI-Generated Image Be Registered as a Trademark?

Potentially, subject to trademark law.

Trademark registration does not necessarily depend on the applicant proving that it personally owns copyright in every graphic element.

The principal trademark questions include matters such as:

distinctiveness,
earlier rights,
prohibited signs,
likelihood of confusion,
and other statutory grounds.

However, if the generated image infringes another person’s copyright, that could create separate problems.

A startup may therefore theoretically possess trademark rights in a sign while facing a separate copyright dispute concerning the underlying design.

Again, different intellectual property rights must be analysed separately.

What About AI-Generated Software Code?

Software startups should apply particular caution.

Generative coding tools may produce:

functions,
libraries,
scripts,
tests,
algorithms,
configuration files,
and entire modules.

Several questions arise.

Is the Generated Code Copyright-Protected?

This depends on whether sufficient human-authored expression exists and whether the code meets copyright requirements.

Can the Startup Claim Exclusive Ownership?

Not necessarily for purely machine-generated code.

Could the Code Reproduce Existing Software?

Potentially.

Could Open-Source Licence Obligations Apply?

Potentially, depending on the source and similarity of the generated material.

Did the Employee Upload Confidential Source Code Into the AI Tool?

This creates separate confidentiality and data-security risks.

A startup should therefore use technical and legal review before placing substantial AI-generated code into production.

AI-Generated Music Creates Similar Problems

A startup may use AI to generate:

background music,
advertising jingles,
game soundtracks,
podcast music,
or video soundtracks.

The same questions arise:

Is there sufficient human authorship?
Does the output reproduce existing compositions?
Does it imitate identifiable recordings?
Are performers’ rights implicated?
Does it reproduce a person’s voice?
What do the platform terms permit?
Can the startup grant exclusive rights to customers?

Music can involve multiple rights simultaneously, including:

composition,
lyrics,
performance,
sound recording,
and personality-related rights.

Therefore, “AI-generated” does not mean “rights-free”.

What About AI Voice Cloning?

Voice cloning illustrates why copyright is not the only relevant legal system.

Suppose an AI system replicates a famous actor’s voice.

The exact output may not necessarily be resolved only through copyright analysis.

Other issues may involve:

personality rights,
personal data,
unfair competition,
misleading endorsement,
contracts,
and potentially sector-specific rules.

A startup generating synthetic voices should therefore avoid asking only:

“Who owns the audio file?”

The more important question may be:

“Did we have the legal right to imitate this person’s voice?”

What About Deepfakes?

Deepfake content can similarly trigger multiple areas of law.

AI-generated video involving a real person may raise:

personality-right claims,
personal data protection,
defamation,
privacy,
copyright,
criminal law,
and AI transparency obligations.

For EU-facing deployments, Article 50 AI Act transparency requirements concerning deepfakes have applied since 2 August 2026.

Ownership of the deepfake file is only one small part of the legal analysis.

Can a Startup Sell AI-Generated Content to Customers?

Potentially, but the contract should be drafted carefully.

Suppose an AI design startup sells generated images to customers and promises:

“You receive exclusive worldwide copyright ownership of every generated image.”

That statement may create contractual risk if the startup cannot establish that copyright actually exists.

A safer contractual structure may distinguish between:

rights the startup legally owns;
rights assigned by human creators;
contractual rights to use the output;
third-party rights;
AI-generated portions;
and limitations inherent in generative systems.

The company should avoid giving absolute warranties that exceed its actual legal position.

“Exclusive AI Content” May Not Truly Be Exclusive

Generative AI systems can sometimes create similar outputs for different users.

Suppose Customer A asks for a particular image.

Customer B later uses a similar prompt and receives a similar image.

If the first image lacks human-authored copyright protection, Customer A may have difficulty claiming conventional exclusive copyright against Customer B.

Even if the AI provider contract promises certain rights, the technical system may not guarantee uniqueness.

Startups selling generative content should therefore be very careful with marketing expressions such as:

“fully exclusive”,
“unique intellectual property”,
“100% copyright ownership”,
or “guaranteed original”.

Can Terms of Use Create Contractual Exclusivity Even Without Copyright?

To some extent, contractual arrangements can regulate the parties.

For example, an AI design agency may agree contractually that it will not intentionally resell a particular commissioned design to another customer.

This can create a contractual obligation between the agency and customer.

However, contract rights are not the same as copyright.

A contractual restriction normally binds the parties to the agreement.

Therefore:

contractual exclusivity ≠ copyright exclusivity.

This distinction should be explained clearly to customers.

Ownership of AI Content in a Startup Should Be Addressed in Employment Policies

Startups should update intellectual property documentation for generative AI.

Employment agreements and AI policies may address:

authorised AI tools,
confidential information,
personal data,
prompts,
AI-generated materials,
employee modifications,
human-created elements,
commercial use,
record keeping,
and third-party infringement risk.

The company should be able to identify which employee contributed what creative elements.

This becomes particularly important during investment due diligence.

AI Content Should Be Documented

If a startup expects a commercially important AI-assisted work to receive copyright protection, it should preserve evidence of human authorship.

This may include:

initial sketches,
drafts,
original written material,
prompts,
iteration history,
editing history,
Photoshop files,
version control,
source files,
human annotations,
and final modifications.

This evidence can demonstrate that the final work was not merely generated through a single automated command.

Why Human Contribution Records Matter

Imagine that a competitor copies a startup’s AI-assisted advertisement.

The startup files a copyright claim.

The competitor argues:

“The entire image was generated by AI and therefore the startup has no copyright.”

The startup may then need to demonstrate:

which parts were created manually;
what artistic choices the employee made;
how the AI output was modified;
and which elements reflect the employee’s individual creativity.

Without records, proving the human contribution may be difficult.

Investors Will Increasingly Examine AI-Generated IP

Venture capital due diligence is already expanding beyond traditional questions such as:

Who owns the source code?
Are the trademarks registered?
Did freelancers assign their rights?

Investors may increasingly ask:

Which company assets were created using generative AI?
Which AI tools were used?
What terms applied?
Was confidential information uploaded?
Are important outputs copyrightable?
Were human modifications documented?
Could outputs infringe third-party rights?
Was AI-generated code reviewed?
Are training datasets lawfully sourced?
Does the company have an AI usage policy?

A startup whose product library consists almost entirely of automatically generated content may therefore face different intellectual property due diligence risks from a company relying on traditional human-authored assets.

AI-Generated Intellectual Property Can Affect Startup Valuation

Consider two design startups.

Startup A

Its entire asset library consists of 500,000 automatically generated images created without meaningful human editing.

Startup B

It uses AI during production, but professional designers:

direct the process,
substantially edit outputs,
create original compositions,
maintain project records,
and integrate AI elements into broader original works.

Startup B may have a stronger argument that at least substantial parts of its commercial library contain protectable human authorship.

This can affect:

licensing value,
enforcement,
acquisition due diligence,
and company valuation.

What Should Startups Do Before Commercially Using AI-Generated Content?

A practical review should consider at least the following.

1. Identify the AI Tool

Which platform or model generated the material?

2. Review the Platform Terms

What rights does the provider grant?

3. Identify Human Contribution

Who made the creative decisions?

4. Preserve Evidence

Maintain drafts and editing history.

5. Review Third-Party Similarity

Does the output resemble known protected works?

6. Review Trademarks

Does the image contain logos or protected signs?

7. Review Personality Rights

Does it depict or imitate real individuals?

8. Review Personal Data

Were personal data used as inputs?

9. Review Commercial Warranties

Do not promise customers more rights than the startup possesses.

10. Review Employee and Freelancer Agreements

Ensure genuine human rights can be exercised or transferred appropriately.

AI Copyright Checklist for Startups

Before treating AI-generated content as company intellectual property, founders should ask:

What AI tool produced the content?
What do the provider’s terms say?
Did a human create the underlying material?
What specific human creative contribution exists?
Was the AI used merely as an editing tool?
Did the AI determine most expressive elements?
Is the prompt itself original?
Was the output substantially modified?
Were multiple outputs creatively selected and arranged?
Can the human contribution be documented?
Was the content produced by an employee?
Was it produced within employment duties?
Does Article 18 apply?
Was the content created by a freelancer?
Is there a written IP agreement?
Are economic rights identified individually where transfer is required?
Does the freelancer actually possess the rights being transferred?
Does the output reproduce existing protected material?
Does it contain third-party trademarks?
Does it depict a real individual?
Does it clone a voice?
Does it contain confidential information?
Was copyrighted source material uploaded to the AI?
Does the generated code resemble third-party code?
Are open-source licensing issues involved?
Is the startup developing its own AI model?
Are training dataset rights documented?
Is the model offered in the European Union?
Do EU AI Act copyright obligations apply?
Are AI transparency obligations applicable?
Are customers promised copyright ownership?
Can those promises actually be fulfilled?
Is exclusivity being promised?
Is the output technically capable of being reproduced for another user?
Can the startup prove the chain of rights during investor due diligence?

If these questions cannot be answered, the startup should avoid treating AI-generated materials as automatically secure intellectual property assets.

Frequently Asked Questions About Copyright in AI-Generated Content

Can artificial intelligence own copyright in Turkey?

Under the current FSEK structure, the author is the person who creates the work, and the legislation does not designate an AI system as an author.

Does the person writing the prompt automatically own the AI output?

Not necessarily. The relevant question is whether the person made sufficient creative contributions to the resulting protected expression.

Can AI-assisted content receive copyright protection?

Potentially. If a human contributes sufficiently original creative expression, those human-authored elements may receive protection even though AI was used during the process.

Does a detailed prompt receive copyright protection?

The prompt itself may potentially qualify as a literary work if it satisfies the statutory criteria. This does not automatically determine copyright ownership of the resulting AI output.

Can a company own AI-generated content created by an employee?

The first question is whether a protected human-authored work exists. Where an employee creates qualifying work while performing employment duties, Article 18 may allow the employer to exercise the relevant rights unless otherwise agreed or indicated by the nature of the work.

Can a freelancer transfer copyright in AI-generated content?

Only rights that actually exist can effectively be transferred. FSEK Article 52 also requires contracts concerning economic rights to be in writing and the relevant rights to be individually identified.

Does paying for an AI subscription mean the user owns copyright?

Not automatically. Subscription terms and statutory copyright ownership are different questions.

Can AI-generated content infringe somebody else’s copyright?

Yes. AI generation does not guarantee that output is free of third-party rights.

Does the EU AI Act decide who owns AI-generated content?

No. The AI Act contains important copyright compliance and transparency requirements, particularly for general-purpose AI providers, but it does not create a general rule that automatically assigns copyright in AI output to the user. Article 53 requires GPAI providers to maintain a copyright compliance policy and publish a training-content summary.

Has Turkey enacted a special rule governing copyright ownership of AI-generated works?

As of 20 August 2026, the FSEK does not contain a specific AI authorship rule. A proposal aimed at updating the FSEK for copyright issues arising in the AI era was submitted in April 2026 but remains in parliamentary commission.

Conclusion: Who Owns the Copyright in AI-Generated Content in Turkey?

The question “Who owns AI-generated content?” appears simple but actually involves several separate legal questions.

Under the current Turkish copyright framework, the starting point is human authorship.

The FSEK defines a protected work through the author’s individual characteristics and states that the author is the person who creates the work.

The statute currently contains no provision establishing an artificial intelligence system as an author.

Nor does it contain a general rule stating:

“The person who enters the prompt automatically owns the output.”

For this reason, purely AI-generated material containing little or no identifiable human creative expression creates substantial uncertainty from a traditional copyright perspective.

The more appropriate analysis asks:

Who made the expressive choices?

If the AI system determined the:

composition,
wording,
artistic details,
structure,
colours,
melody,
characters,
or other expressive elements

with minimal human control, establishing conventional human authorship may be difficult.

If, however, the AI system functioned merely as one tool within a wider creative process and a human:

created substantial original material,
determined composition,
modified outputs,
selected and arranged elements creatively,
rewrote text,
edited imagery,
or otherwise made meaningful expressive decisions,

those human-created elements may have a much stronger claim to copyright protection.

Therefore, the most important distinction for startups is:

AI-generated content versus AI-assisted human creativity.

That distinction also affects employee ownership.

Where an employee creates a qualifying human-authored work while carrying out employment duties, Article 18 can give the employer authority to exercise the relevant rights under the statutory conditions.

Where freelancers are used, contractual documentation becomes particularly important. Economic-right transfers under the FSEK must be made in writing and the rights concerned must be individually identified.

But no employment agreement or freelancer contract can solve the threshold question by itself:

Does a copyright-protected work exist in the first place?

Startups should also remember that copyright ownership is only one half of the AI intellectual property problem.

An output may have uncertain copyright protection while simultaneously infringing an existing third-party work.

Accordingly, companies should not assume:

“Because the AI created it, it must be safe to use.”

Commercial AI outputs should be reviewed for:

copyrighted material,
trademarks,
software licences,
personality rights,
confidential information,
voice imitation,
and other third-party rights.

For startups developing their own artificial intelligence models, the legal analysis becomes even broader.

The company must consider not only output ownership but also:

training datasets,
reproduction rights,
licences,
text and data mining rules,
rights reservations,
data protection,
and international AI regulation.

For Turkish startups offering general-purpose AI models in the European Union, Article 53 of the EU AI Act is particularly important. It requires GPAI providers to implement a policy for compliance with EU copyright law and to publish a sufficiently detailed summary concerning training content.

From 2 August 2026, separate EU AI Act transparency obligations also apply to certain AI-generated and manipulated content.

However, none of these obligations should be confused with copyright ownership.

A transparency label saying:

“Generated with AI”

does not decide whether copyright exists.

A platform contract saying:

“The user owns the output”

does not necessarily create copyright where copyright law does not recognise a protectable work.

A payment receipt showing:

“Premium AI subscription”

does not establish authorship.

And a prompt saying:

“Create something original”

does not guarantee that the resulting output is free from third-party rights.

For founders, the safest approach is therefore to create an AI intellectual property governance system.

Commercially important AI-assisted assets should be documented.

The startup should preserve:

human drafts,
prompts,
editing history,
source files,
design decisions,
model information,
platform terms,
and records of human modifications.

Employees and freelancers should be required to disclose material use of generative AI in important company assets.

AI-generated software should be technically and legally reviewed.

AI-generated logos should undergo trademark clearance.

AI-generated marketing content should be checked for third-party works and personality rights.

Customer contracts should avoid guaranteeing copyright exclusivity where the startup cannot confidently establish it.

This approach becomes especially important when the startup seeks investment or prepares for an exit.

An investor examining an AI-heavy startup will increasingly ask:

Does the company actually own the intellectual property it claims to own?

For traditional startups, answering that question may require reviewing employee and freelancer agreements.

For AI startups, the investigation can go further:

Which parts were created by humans?

Which parts were produced entirely by AI?

What rights do the AI provider’s terms actually grant?

Can competitors lawfully reproduce the same output?

Was copyrighted content used during training?

Could important outputs infringe third-party rights?

Can the company prove a legally defensible chain of title?

These questions directly affect the value of the startup.

The practical principle is therefore:

Do not assume that every commercially valuable AI output is automatically a copyright asset.

Instead, determine where the human creativity exists, document that creativity, secure the relevant rights and separately investigate third-party infringement risk.

As artificial intelligence becomes a normal part of commercial creativity, copyright law will increasingly focus not merely on whether AI was used, but on how it was used and what creative role remained with the human.

For startups operating in Turkey, that distinction is currently the key to understanding copyright ownership in AI-generated content.

Categories:

Genel

Who Owns the Copyright in AI-Generated Content Under Turkish Law?