Who is the Real Owner of Artifacts in the Age of AI?

Who is the Real Owner of Artifacts in the Age of AI? The rise of generative artificial intelligence (AI) systems, particularly platforms like ChatGPT, Midjourney, and DALL-E2, has caused a paradigm shift in the creative industries. These technologies, with their capacity to produce text, visuals, music, and even software code with human-like proficiency, are shaking the foundations of copyright law. This legal framework, designed over centuries to protect works originating from the human mind, is now confronted with content produced by autonomous or semi-autonomous machines. A system built on the philosophical and legal foundations of the 18th and 19th centuries is in direct conflict with the technological reality of the 21st. This situation is reshaping debates about the existence and scope of copyright protection, making them dependent not only on the nature of the content but also on the level of human involvement in the production process.

The primary purpose of this report is to seek an answer to one of modern law’s most complex questions: Who owns the copyright to content prepared by artificial intelligence? To answer this, we will first analyze the status of copyright law’s foundational concepts, such as “author” and “originality,” in the face of AI. We will then comparatively examine the current legal frameworks, landmark cases, and regulatory approaches in the United States, the European Union, the United Kingdom, and China—nations leading the legal and political debates in this field. Finally, we will assess the current situation in Turkish law, discuss potential ownership models, and offer a strategic evaluation of the economic, ethical, and legal future of this technological revolution. This analysis aims to provide a roadmap for a wide range of stakeholders, from lawyers and technology leaders to policymakers and artists.

Chapter 1: The Cornerstones of Copyright: Re-evaluating “Author” and “Originality”

1.1 The Traditional Requirements for Copyright Protection

The protective mechanism of copyright law has historically been built on specific fundamental principles centered on human creativity. The inclusion of artificial intelligence in this domain forces us to question the validity and applicability of these principles.

Author: The almost universally accepted cornerstone of copyright law is the identification of the “author” with a human being, a natural person. This principle is at the heart of the legal struggle. For example, in Turkish law, the Law on Intellectual and Artistic Works (FSEK) No. 5846 explicitly defines the author as “the person who creates the work,” and it is generally accepted that this refers to a human. Similarly, U.S. copyright law, derived from the U.S. Constitution’s reference to “Authors” and “Writings,” has been consistently interpreted by courts to exclude non-human creators. This human-centric approach leads to the conclusion that a machine, lacking legal personality, cannot be an author.

Originality and Personal Touch: For a work to benefit from copyright protection, it must be “original.” Originality does not mean it has never existed before (novelty); rather, it means the work originates from the author and contains a minimal degree of creativity. In Turkish law, this concept appears as the requirement for the work to “bear the author’s personal touch” (hususiyet). This means the work must carry the personal, intellectual, and creative stamp of its author. Applying this requirement to content produced by a machine lacking personality, consciousness, or subjective experience is extremely difficult, as a machine cannot exhibit a “personal touch.”

Intellectual Effort: Copyright protects the tangible expression of an abstract idea, not the idea itself. It is essential that this expression is the product of the author’s own intellectual effort. Legal systems do not deem mere labor or effort sufficient for protection; the “sweat of the brow” doctrine has been generally rejected. This means the time and effort spent writing a prompt for an AI model may not be sufficient on its own for copyright protection if that effort does not translate into a creative expression in the final output.

1.2 The Role of AI in the Creative Process and the Concept of “Creative Control”

AI’s role in the creative process spans a wide spectrum, from a simple tool to an autonomous creator. The critical factor determining the existence and scope of copyright protection is where on this spectrum the work falls and the degree of “creative control” retained by the human.

AI as a Tool: When AI functions as a sophisticated tool to help a human realize their creative vision—much like a photographer’s camera or a graphic designer’s Photoshop—the human contribution to the resulting work is generally protectable. For example, if a photographer uses AI to remove an unwanted object from a photo they took, this does not negate the human authorship of the photograph. In this scenario, the ultimate creative control over the final expression rests with the human.

AI as a Collaborator: This is where the greatest legal uncertainty lies. A human gives the AI model specific prompts, and the AI generates content based on these commands. The key question here is: Who is responsible for the “expressive elements” of the work? Regulatory bodies like the U.S. Copyright Office (USCO) argue that if the AI “determines the expressive elements of its output,” this material cannot be considered a human work.

AI as an Autonomous Creator: In scenarios where an AI system produces a work with minimal or no human intervention (as in the Thaler case), the resulting product is not eligible for copyright protection under current U.S. and EU law. The primary reason is the absence of a human author.

Creative Control: At the center of these discussions, “creative control” is emerging as a new legal test. The focus is not on the abstract idea expressed in the prompt, but on the control over the expression embodied in the output. The USCO likens a user writing a prompt to a client giving general instructions to an artist; in that case, the client is not considered the author. Therefore, the degree of human control over the selection, arrangement, and modification of the AI output has become the main factor in determining copyright ownership.

At this juncture, a subtle shift in legal discourse is visible. While “human authorship” remains a legal necessity, the test applied in practice is evolving into an inquiry that questions the degree of human creative control. Traditional copyright laws were built on a binary logic of whether a human author existed or not. Generative AI, however, has created a spectrum of human involvement, from being used as a simple tool to providing a single prompt. The “human author” test is insufficient to resolve this complexity. Regulators like the USCO have responded not by focusing on the author’s identity, but on the function they perform: “To what extent did a human have creative control over the work’s expression?” This approach represents a shift from a rigid status inquiry (“Is the creator a human?”) to a more flexible, process-based one (“Did a human control the creative process?”). This transformation explains why partial copyright was granted in cases like Zarya of the Dawn: the human controlled the text and page layout, but not the pixel-by-pixel creation of the images.

Chapter 2: The United States Approach: The Fortress of “Human Authorship”

The United States has taken the clearest and most rigid stance in the debate over artificial intelligence and copyright. Guidance from the U.S. Copyright Office (USCO) and court rulings have established the “human authorship” principle as an indispensable prerequisite for copyright protection.

2.1 The Doctrine of the U.S. Copyright Office (USCO)

The USCO, with its AI initiative launched in March 2023, has clearly articulated its policy in this area.

The “Human Authorship” Principle: The USCO’s official guidelines state unequivocally that copyright protection is limited to works of human creation. This principle forms the basis of all its subsequent decisions and reports. If the traditional elements of authorship in a work are produced by a machine, that work lacks human authorship and cannot be registered.

Disclosure Requirement: Applicants for copyright registration must disclose any AI-generated content in their work that is more than de minimis (trivial) and disclaim these portions from their protection claim. Failure to comply with this rule can lead to the cancellation of registration. Indeed, in the Zarya of the Dawn case, the initial registration was re-evaluated and limited after the use of AI was subsequently discovered.

Protection of Human Contributions: The USCO explicitly states that human-created components within a larger work containing AI-generated material can be copyrighted. This protection covers the creative selection, coordination, or arrangement of AI outputs, or significant creative modifications made to those outputs.

2.2 In-Depth Analysis of Landmark Cases

The legal framework in the U.S. has been shaped by a series of key cases.

Thaler v. Perlmutter (Autonomous Machine): Computer scientist Dr. Stephen Thaler attempted to register an image “autonomously” generated by his AI system, the “Creativity Machine.” His application was rejected by the USCO, and this decision was upheld by federal courts. The courts affirmed that “human authorship is a foundational requirement of copyright.” This case definitively established that fully autonomous AI products cannot receive copyright protection in the U.S.

Zarya of the Dawn (Hybrid Work): Kristina Kashtanova’s graphic novel, which included images generated by Midjourney, received limited copyright protection from the USCO. While the USCO protected the human-written text and the creative arrangement of the text and images, it excluded the AI-generated images themselves from copyright protection. This decision established the principle that a work can be “dissected” to protect only its human-created parts.

Théâtre D’opéra Spatial and SURYAST (The Limits of Prompting): These two decisions further clarified the USCO’s stance. In Théâtre D’opéra Spatial, registration was refused because the human modifications to an AI-generated image were deemed too minor and insignificant. In the SURYAST case, the applicant used an AI tool to merge a photograph he took with the style of Van Gogh’s Starry Night. The USCO rejected the copyright claim, determining that the AI decided how to combine the two images, meaning there was insufficient human creative control.

2.3 Training Data and the “Fair Use” Doctrine

The other side of the copyright debate concerns the legality of training AI models on copyrighted materials.

The Core Conflict: AI companies argue that copying billions of copyrighted data points scraped from the internet for training purposes is a “transformative” use, similar to the Authors Guild v. Google (Google Books) case, and is legal under the “fair use” doctrine.

Lawsuits from Rights Holders: Authors, artists, and publishers have filed dozens of lawsuits, claiming this practice constitutes mass copyright infringement. Rights holders argue this use is commercial, uses the entirety of the works, and directly harms the market by creating products that compete with the original works.

Thomson Reuters v. ROSS Intelligence: This case, decided in February 2025, was a significant victory for rights holders. A federal court rejected the “fair use” defense of an AI company that trained its model on copyrighted legal summaries. The court found the use was commercial and directly competed with the original product. Although the AI in this case was not generative, the decision is seen as a sign that courts may look skeptically upon fair use claims in commercial contexts.

U.S. copyright law is creating a significant legal paradox. On one hand, it refuses to grant copyright protection to AI outputs, citing a lack of human authorship. On the other hand, it faces intense legal challenges that could potentially restrict or outlaw the use of copyrighted data as inputs for AI training. The USCO and the courts have drawn a clear line on outputs: no human author, no copyright. This effectively places fully AI-generated content into the public domain. At the same time, the massive datasets required for AI models to function are largely composed of copyrighted materials. Rights holders are suing, arguing this data collection is infringement and targeting the “fair use” defense, which is the AI industry’s primary shield. If courts rule against fair use for training data, AI companies will face either massive damages or expensive licensing agreements. This puts the AI industry in a bind: the products of AI may be unprotected and free for all to use, while the process of creating those products could become legally and financially unfeasible. This paradox creates extreme uncertainty that threatens to stifle innovation, not from a lack of protection for outputs, but from the legal risk associated with inputs.

Chapter 3: The European Union’s Regulatory Framework: The AI Act and Its Indirect Impact on Copyright

The European Union is approaching artificial intelligence not by directly rewriting copyright laws, but by establishing a comprehensive regulatory framework. The AI Act, approved in March 2024 and a first of its kind, contains provisions that will indirectly but powerfully affect copyright issues.

3.1 The EU AI Act: A Risk and Transparency-Focused Approach

The AI Act aims to create a safe, transparent, and ethical environment for the development and use of AI systems, rather than to reform copyright.

Risk-Based Framework: The law classifies AI systems according to risk levels (unacceptable, high, general-purpose, minimal) and imposes stricter rules for high-risk systems. This approach aims to ensure regulation is proportional and does not stifle innovation.

Transparency Obligations: The most critical provisions for copyright holders relate to transparency. Providers of general-purpose AI models, such as ChatGPT, are subject to the following obligations:

Clearly state that the content they produce is AI-generated.
Publish “sufficiently detailed summaries” of the copyrighted data they used to train their models.

3.2 Text and Data Mining (TDM) and the Control of Rights Holders

The obligations introduced by the AI Act must be evaluated in conjunction with the EU’s previous copyright regulations.

The DSM Directive (2019/790): The Directive on Copyright in the Digital Single Market introduced an exception for text and data mining (TDM) for scientific research and, under certain conditions, for commercial purposes. TDM forms the basis of the AI model training process.

The “Opt-Out” Mechanism: One of the most important elements of the Directive is that it grants rights holders the right to prevent their works from being used for TDM, especially for commercial purposes. Rights holders can “opt-out” of this exception by issuing a “reservation of rights” in machine-readable formats. The AI Act strengthens this mechanism by obligating AI providers to respect these reservations.

Extraterritorial Effect: The AI Act applies to all AI systems offering services in the EU market, regardless of where the provider is located. This means that U.S.-based or other non-EU AI companies will have to comply with these transparency and “opt-out” rules if they wish to operate in the EU. This has the potential to turn the EU’s regulatory standards into a de facto global norm.

The EU is avoiding a direct answer to the “who is the author?” question. Instead, it is using its regulatory power through the AI Act to shift the balance of power between tech companies and rights holders. In an environment where AI models are “black boxes” and it is difficult for rights holders to prove their works were used in training, the transparency requirement imposed by the EU eliminates this information asymmetry. The AI Act forces AI companies to open this black box and publish summaries of their training data. An artist or publisher armed with this information is in a stronger position to file an infringement lawsuit under existing copyright laws or to demand a license. This approach is a pragmatic strategy that empowers rights holders and encourages the development of licensing markets, rather than getting bogged down in a philosophical debate over AI authorship. This is market-shaping regulation rather than fundamental legal reform.

Chapter 4: The United Kingdom’s Quest: A Path Between Tradition and the Future

The United Kingdom is on a complex quest regarding artificial intelligence and copyright. On one hand, it seeks to create an innovation-friendly environment to become a global AI hub, while on the other, it is under pressure to protect its powerful creative industries. This has led to significant uncertainty in the legal framework and delays in policymaking.

4.1 The “Computer-Generated Works” Provision

The UK’s Copyright, Designs and Patents Act 1988 (CDPA) contains a rare provision stating that the author of a “computer-generated work” is “the person by whom the arrangements necessary for the creation of the work are undertaken.” While considered forward-thinking at the time, this provision is inadequate in the face of the complexity of modern generative AI systems. The government has acknowledged this provision is outdated and is considering its removal.

4.2 Current Legal Uncertainty and Delayed Legislation

Although the UK government has acknowledged the need for legal clarity, it has postponed a comprehensive AI bill until at least 2026. This delay creates a significant climate of uncertainty for both the tech sector and the creative industries. Companies and artists are forced to operate without clear guidance on the legal status of AI training and outputs, which risks negatively impacting investment and innovation.

4.3 The Search for New Regulations

To address this uncertainty, the government has initiated a series of consultation processes. The proposed new framework largely mirrors the EU’s approach. It discusses introducing a new TDM exception that includes a clear “opt-out” mechanism for rights holders. Also on the agenda are issues like the labeling of AI outputs and the establishment of a collective licensing system to ensure authors are compensated for the use of their works in training data. These proposals aim to ensure artists can earn a fair income even in situations where they cannot negotiate individually with tech companies.

In the post-Brexit era, the UK faces a difficult dilemma. On one hand, it wants to position itself as a global AI power with a light-touch regulatory environment to attract investment. On the other, it recognizes the need to protect its strong creative industries and finds itself drawn toward the EU’s regulatory model (a TDM exception with an opt-out) to solve the training data problem. This has led to policy indecision. Post-Brexit, the UK has the freedom to diverge from EU law, and the initial goal was to create a more innovation-friendly, less regulated environment for AI. However, the issue of copyright training data presents a major obstacle; the creative industries are lobbying hard against uncontrolled data scraping. The government’s proposed solution—a TDM exception with an opt-out mechanism—is nearly identical to the EU’s DSM Directive model. This reveals the existence of a regulatory gravitational pull. Despite the desire for divergence, the EU’s solution is seen as a viable compromise that balances the interests of the tech and creative sectors. The legislative delay reflects the difficulty of resolving this tension: the desire for a unique UK approach versus the practical appeal of the established EU model.

Chapter 5: China’s Pragmatic Step: User-Centric Copyright Protection

While legal and philosophical debates continue in the West, China has taken a pragmatic and remarkably different path regarding the copyright status of AI-generated content. Decisions from the Beijing Internet Court have the potential to set a global precedent.

5.1 Landmark Rulings of the Beijing Internet Court

In a landmark decision known as Li v. Liu, issued in November 2023, the Beijing Internet Court granted copyright protection to an AI-generated image and ruled that the author was the userwho wrote the prompts. This decision stands in stark contrast to the approaches in the U.S. and EU, where AI outputs are denied copyright protection.

5.2 The Criteria of “Intellectual Contribution” and “Personalized Expression”

The court’s decision departs from the Western “human authorship” doctrine and focuses on the user’s creative process. The court ruled that the user’s process of selecting the AI model, crafting detailed prompts, adjusting parameters, and refining the resulting outputs met the “intellectual achievement” required by Chinese Copyright Law. This process was deemed to reflect the user’s “personalized expression” and “aesthetic choice.” According to the court, the user did not merely state an idea but made a series of creative choices that shaped the final appearance of the work.

5.3 Divergence from the U.S. Approach

The Beijing Internet Court agrees with the West that the AI model itself cannot be an author. However, the fundamental difference lies in how prompts are evaluated. While the USCO views prompts as unprotectable ideas, the Chinese court treated them as creative inputs that directly shape the final, protectable expression. This approach shifts the focus from the machine’s autonomy to the human’s creative labor in guiding the machine. A more recent decision, affirmed in September 2025, reinforced this approach, stating that the user claiming copyright has the burden of proof to explain their creative thought, the content of their input prompts, and the process of selecting/modifying the generated content with evidence.

China’s approach is not just a different legal interpretation; it is a strategic economic policy. By granting copyright to the user, the court is encouraging and legitimizing new and emerging fields like “prompt engineering” and AI-assisted content creation. This nurtures a domestic ecosystem for generative AI applications and aims to position China as a leader in the commercial exploitation of this technology. The Western approach, by leaving AI outputs largely in the public domain, discourages commercial investment in creating high-quality, AI-assisted works; why spend time crafting the perfect prompt if the resulting work is free for anyone to copy? China’s model does the opposite: it says, “Your skill in using an AI tool to create something original is valuable and will be legally protected.” This defines the skilled AI user as a new class of “author” and encourages the development of professional prompt engineering and marketplaces for user-owned, AI-generated content. Legal framework provides a clear path to ownership and monetization that is missing in the West, potentially giving Chinese creative tech companies a competitive advantage. This is industrial policy enacted through copyright law.

Chapter 6: The Current Situation and Future Prospects in Turkish Law: The Silence of FSEK

In Turkey, the copyright status of AI-generated content exists in a legal vacuum. The Law on Intellectual and Artistic Works (FSEK) No. 5846 does not contain provisions specific to this new technological reality. Therefore, the current situation and potential future developments depend heavily on the interpretation of existing law and the influence of international trends.

6.1 FSEK’s Existing Definitions

FSEK’s current definitions suggest that content autonomously generated by AI will fall outside copyright protection.

“Author”: FSEK Article 1/B-b defines the author as “the person who creates the work.” There is a strong consensus in Turkish legal doctrine that the term “person” refers to a natural person (a human), not machines, which are legal objects. Even legal entities (like corporations) cannot be authors, as they cannot engage in creative intellectual work; they can only hold the economic rights to a work. This interpretation is the biggest obstacle to recognizing AI as an author.

“Work” and “Personal Touch” (Hususiyet): According to FSEK Article 1/B-a, for an intellectual product to be considered a “work,” it must “bear the author’s personal touch.” “Personal touch” means the work carries traces of the creator’s personality, intellectual and aesthetic choices—that is, it exhibits an original character. It does not seem possible, within the current legal understanding, for content produced by AI, which lacks consciousness and a personal perspective, to meet this requirement. Therefore, a fully AI-generated output may not qualify as a “work” under FSEK.

6.2 The Search for Solutions Through Interpretation and Analogy

In the face of the current legal gap, disputes in this area will be resolved by courts through the interpretation of existing principles and by analogy. It is highly likely that Turkish courts will follow the dominant international trend (especially in the U.S. and EU) and look for the presence of meaningful human creative intervention for a work to be protectable. In this scenario, AI will be evaluated as a “tool” under human control, not as a “creator” at the center of the creative process.

6.3 Possible Scenarios

Turkey has several possible paths forward. The first is to maintain the status quo and leave the issue to the development of case law. The second is to introduce specific provisions for AI-generated content through an amendment to FSEK. A third possibility is to create a new, sui generis (of its own kind) protection category for such content, separate from copyright. This new category could grant more limited rights than traditional copyright, thereby both encouraging innovation and providing some protection for creative labor.

Due to its silence and its strong emphasis on the concept of “hususiyet” (personal touch), Turkish law is de facto aligned with the restrictive US/EU model. However, this is an alignment by default rather than a conscious policy choice. FSEK was written long before generative AI, and its core concepts are deeply tied to the romantic notion of human authorship. Without new legislation, courts applying FSEK will almost certainly conclude that autonomous AI works are unprotected. This places Turkey in the same camp as the US and EU. However, in the US and EU, this outcome is the product of active, ongoing debates, policy statements, and new regulations (like the AI Act). In Turkey, it is the result of legislative inaction. This inaction creates a risk. While other countries are actively designing legal frameworks to either restrict or incentivize AI innovation (e.g., the EU AI Act, China’s court rulings), Turkey’s legal environment remains uncertain, which could deter investment and leave both creators and tech companies in a state of legal limbo.

Chapter 7: Ownership Models, Economic and Ethical Dimensions

The legal uncertainty surrounding the copyright ownership of AI-generated content is not just a theoretical debate; it is a practical problem with profound economic and ethical consequences for the creative and technology industries. This chapter examines potential ownership models, their advantages and disadvantages, and the impact of the issue on stakeholders.

7.1 Comparison of Potential Ownership Models

Three primary ownership models have emerged in legal discussions:

User as Author: Pioneered by China, this model recognizes the user who guides the AI and enters the creative prompts as the author of the work.

Advantages: It incentivizes skilled AI use and the “prompt engineering” market. It rewards the human’s creative labor in directing the tool and offers a clear path to ownership for content creation.
Disadvantages: It could lead to the market being flooded with low-effort, copyrighted content. It blurs the line between giving instructions and creating.

Developer as Author: In this model, the company or person who developed the AI system owns all content the system produces.

Advantages: It rewards the massive investment required to build AI models and draws a clear line of ownership.
Disadvantages: It rests on a weak legal foundation, as the developer has no intent or direct involvement in creating a specific output. It could lead to a monopolistic structure where a few large tech companies own all AI-generated cultural content.

Public Domain: This is the default model in the U.S. for fully AI-generated content. In this model, the content belongs to no one and is free for all to use.

Advantages: It encourages free access and reuse of content, preventing the “lock-up” of culture by AI.
Disadvantages: It devalues creative labor. It discourages the creation of high-quality, AI-assisted works because they cannot be monetized or protected against copying.

7.2 Impacts on Artists and Creative Industries

Economic Impacts: The biggest concern for artists, writers, and designers is that AI will take their jobs by offering a cheaper, faster alternative. AI’s ability to instantly mimic an artist’s style risks devaluing the skill and creativity that humans spend years developing.

Ethical and Intellectual Property Concerns: The use of artists’ works in AI training data without their permission or compensation is seen by many creators as “mass theft.” AI outputs can create unauthorized derivative works by imitating an artist’s unique style, diluting the artist’s brand. This leads to artists losing control over their own work and the exploitation of their labor.

7.3 Impacts on the Technology Industry

Legal Risks: The technology industry, especially AI developers, faces massive potential liability from copyright infringement lawsuits related to training data.

Innovation Incentives: Legal uncertainty deters investment. Even a restrictive legal framework is often preferable to uncertainty. The outcome of the “fair use” debate constitutes the single greatest legal threat to the current business model of generative AI companies.

The U.S. approach of defaulting AI-generated works to the public domain, while seemingly democratic, may paradoxically harm human creators and concentrate market power. USCO’s stance means an image a user creates with AI is likely in the public domain. This prevents the user from protecting their own creation. But it also means that large commercial entities can take that public domain image and use it for their own profit, paying nothing to the original prompt author. This devalues the skill of prompt engineering and makes it difficult for individual creators to build a business around AI-assisted art. Meanwhile, large corporations with their own legal teams and resources are better positioned to exploit this vast new pool of “free” content. Furthermore, the legal risk associated with training data favors large companies that can afford to license massive datasets or absorb potential legal damages. The result is that the combination of unprotected outputs and legally risky inputs may create a market where only the biggest players can innovate safely, and individual human creators are squeezed from both ends: they cannot protect their outputs and face the uncompensated use of their original works as inputs.

Chapter 8: Conclusion and Strategic Assessment

The copyright status of AI-generated content is a complex, multi-layered problem that is fundamentally challenging global legal systems. The analysis throughout this report shows that rather than a single global solution, diverse approaches are emerging based on different philosophical, economic, and legal priorities.

8.1 Summary of Global Approaches and Key Divergences

The current global landscape is shaping up around two main axes. On one side is the human-centric model, led by the U.S. and the EU, which strictly ties copyright protection to human creativity. This model aims to protect the traditional author and leaves autonomous AI outputs outside the scope of protection. On the other side is the user-centric model adopted by China, which protects the modern creator who skillfully uses AI as a tool. This model deems the “intellectual contribution” made through prompts and parameters sufficient for copyright protection. Countries like the United Kingdom and Turkey are, for now, trying to find their own way between these two poles, amid legal uncertainties and calls for reform.

8.2 Looking to the Future: The Evolution of Law and New Balances

The future of copyright law will continue to evolve in line with technological developments and societal needs. In the near future, the proliferation of licensing markets for AI training data is expected. The EU’s transparency requirement may serve as a catalyst to accelerate the formation of these markets. It is also possible that some countries will establish sui generisrights for AI-generated content, providing more limited protection than traditional copyright. Finally, private law contracts, such as the Terms of Service of AI platforms, will play an increasingly important role in determining ownership and usage rights in areas where legal gaps exist.

8.3 Strategic Recommendations

In this complex and rapidly changing environment, it is critical for stakeholders to adopt a proactive and informed approach.

For Creators: Digitally documenting works (e.g., using timestamps) and establishing proof of ownership will become increasingly important. If they do not want their works used in AI training, they must actively use “opt-out” mechanisms, as available in the EU, and mark their works accordingly. Those who use AI in their creative processes should keep detailed records of their prompts, edits, and selection processes to be able to prove the degree of human contribution and creative control.
For Developers: Proactively complying with transparency obligations imposed by regulations like the EU AI Act will reduce long-term legal risks. Using ethically and legally cleared data sources is the safest way to avoid copyright lawsuits. Developing fair and scalable licensing models with rights holders will be unavoidable for a sustainable business model.
For Policymakers: In countries with legal vacuums, like Turkey, policymakers must take active steps to fill this gap by pursuing a balanced policy that encourages innovation while protecting the rights of creators. Closely monitoring international norms and landmark cases is essential to avoid falling behind in global competition.

The following table summarizes the approaches in the key jurisdictions analyzed, offering a clear comparison.

Table 1: Comparison of International Approaches to AI Copyright

Jurisdiction	Core Principle	Copyrightability of Fully AI-Generated Content	Potential Rights Holder	Landmark Case / Law
United States	Strict “Human Authorship”	No, not protected (Considered public domain)	Owner of human contributions only (e.g., text, arrangement)	Thaler v. Perlmutter, Zarya of the Dawn, USCO Guidance
European Union	“Author’s Own Intellectual Creation”	No, not protected (Requires human author)	Human creator using AI as a tool	EU AI Act (Indirect impact), DSM Directive
United Kingdom	Unclear (Legislation Pending)	“Computer-generated work” provision exists, but status is uncertain	Likely the person who made arrangements (under old law), but legal reform is expected	CDPA 1988, Government Consultations
China	“Intellectual Contribution” & “Personalized Expression”	Yes, protectable	The user who guides the AI and provides creative input	Li v. Liu (Beijing Internet Court)
Turkey	“Author’s Personal Touch” (Hususiyet)	No, not protected (No regulation, based on current interpretation)	Human creator using AI as a tool (Based on interpretation)	5846 sayılı FSEK

10 Critical Questions in AI Law

Our Latest Articles

« Previous Post