Introduction and Research Framing

The rapid proliferation of Artificial Intelligence (AI) technologies has ushered in a transformative era, profoundly reshaping industries, economies, and societal structures. Among the most intriguing and challenging impacts is AI’s burgeoning role in creative endeavors. From Midjourney-generated art winning awards at exhibitions to AI-mimicked voices of renowned singers sparking copyright disputes, AI is permeating and reshaping the creative landscape at an unprecedented pace. These groundbreaking advancements, while unlocking immense possibilities for human creativity and expression, simultaneously introduce complex ethical, legal, and philosophical dilemmas, particularly concerning the fundamental concepts of authorship and accountability. The traditional paradigms, meticulously constructed over centuries to delineate human creative input, ownership, and responsibility, now face unprecedented strain. Who is the author when an AI co-creates, or even autonomously generates, a piece of art? Where does moral agency reside when an AI system produces content that is infringing, harmful, or ethically questionable? These questions, far from being mere academic curiosities, constitute urgent practical concerns for artists, technologists, legal practitioners, policymakers, and the broader public. They underscore a critical need for a normative study that can provide clarity and propose actionable frameworks in this nascent and rapidly evolving landscape.

This normative study aims to address these critical gaps by clarifying where moral agency sits—whether with the AI tool itself, the human user, or the platform/developer—and subsequently proposing a workable authorship model for AI-assisted creative works. Our objective is to navigate the intricate interplay of technology, creativity, and responsibility, offering a robust framework that is philosophically rigorous yet pragmatically relevant to contemporary practice. The “normative” nature of this research signifies that it not only seeks to analyze the current dilemmas surrounding authorship and accountability in AI-assisted creative works but also endeavors to construct a forward-looking and prescriptive theoretical framework and practical model. This framework is intended to provide a solid philosophical and practical foundation for future legal, policy, and industry standard-setting. The interdisciplinary nature of this research is paramount, drawing extensively from diverse fields to construct a comprehensive understanding. We will delve into the philosophy of technology to explore the nature of AI as a creative tool and its implications for human agency and intentionality. Legal studies, particularly intellectual property law (copyright, patent, trade secret) and tort law (liability for harm or infringement), will provide the necessary framework for analyzing existing legal doctrines and their applicability, or lack thereof, to AI-generated content. Ethical considerations, encompassing moral philosophy, questions of responsibility, and distributive justice, will guide our inquiry into assigning accountability in complex AI ecosystems. Finally, insights from creative industry studies will inform our understanding of the evolving practices, concerns, and needs of various creative guilds as they integrate AI into their workflows. This study is not merely a compilation of disciplinary knowledge but strives for deep interdisciplinary integration and dialogue. For instance, we will explore how theories of moral agency from philosophy can provide ethical justification for legal attribution of responsibility, and conversely, how challenges in legal practice can inform and deepen our understanding of AI ethics.

To ensure conceptual precision throughout this study, it is crucial to define key terms, while also acknowledging their dynamic and often contentious nature in the context of AI. AI-assisted creative works refer to any artistic, literary, musical, or other creative output where Artificial Intelligence technologies play a significant role in the conception, generation, modification, or refinement of the work. This spectrum ranges from AI as a mere assistive tool (e.g., AI-powered editing software) to AI as a generative engine (e.g., large language models producing text, image generators creating visuals), and even potentially autonomous AI systems; a core challenge lies in defining the AI’s role and its impact on traditional creative paradigms. Authorship traditionally denotes the individual or entity primarily responsible for the creation of a work, holding rights and responsibilities associated with it. In the context of AI, this concept’s boundaries become fluid, necessitating a re-evaluation of elements traditionally emphasized, such as “originality” and “human intellectual creation.” Accountability refers to the obligation or willingness to accept responsibility for one’s actions, and in this context, for the outcomes, positive or negative, of AI-assisted creative processes. This includes legal liability for infringement or harm, as well as ethical responsibility for biased or problematic outputs; its complexity stems from the fact that the responsible entity may no longer be a single individual but a distributed system involving multiple participants. Moral agency is a central philosophical concept referring to an individual’s or entity’s capacity to make moral judgments based on notions of right and wrong and to be held responsible for those judgments. This study will critically examine whether, and to what extent, AI systems possess moral agency, and the profound implications this has for the attribution of responsibility. Finally, creative guilds encompass professional associations, unions, and communities representing specific artistic and creative disciplines (e.g., visual artists, musicians, writers, designers, filmmakers). Understanding their perspectives and evolving practices is vital for developing practical and acceptable solutions.

This study will adopt a comparative jurisdictional approach, focusing on the European Union (EU), the United States (US), and Japan. These jurisdictions represent distinct legal traditions, policy approaches, and cultural attitudes towards technology and intellectual property, offering a rich comparative landscape. The EU, with its emphasis on human-centric AI and robust data protection regulations (e.g., GDPR, the forthcoming AI Act), provides a perspective rooted in fundamental rights and ethical governance, positioning it as a global bellwether for AI governance and rights protection. The US, characterized by its common law system, strong emphasis on innovation, and evolving copyright doctrines, offers insights into market-driven approaches and the challenges of adapting existing intellectual property frameworks, with recent US Copyright Office rulings on AI-generated works and Silicon Valley’s role as an AI innovation hub making it a key case study for the tension between technological development and legal adaptability. Japan, known for its proactive stance on AI development and unique cultural perspectives on technology and creativity, presents an interesting case study in balancing technological advancement with societal values, offering a valuable reference point distinct from Western contexts, particularly given its leadership in robotics ethics and creative industries like anime and gaming, and its unique cultural view of AI as a ‘tool’. Examining these diverse contexts will reveal common challenges, divergent interpretations, and potential best practices, thereby strengthening the generalizability and applicability of our proposed authorship model.

Ultimately, this study seeks to answer the overarching research question: What scope and questions keep a normative study on authorship and accountability in AI-assisted creative works philosophically rigorous yet relevant to practice, across diverse jurisdictions and creative guilds? By meticulously defining terms, outlining our interdisciplinary approach, and specifying our jurisdictional focus, we lay the groundwork for a comprehensive exploration of this pressing issue. The final objective of this research is to provide legislative guidance for policymakers, codes of conduct for creative industries, and rights protection for artists, thereby fostering the healthy and responsible development of AI technology in the creative domain. The subsequent sections will delve into the conceptual foundations, analyze comparative legal landscapes, integrate stakeholder perspectives, and ultimately propose a workable authorship model, ensuring that our inquiry remains grounded in both theoretical depth and practical applicability.

Reshaping Creation: Interdisciplinary Conceptual Foundations for Authorship and Accountability in AI-Assisted Creative Works

To rigorously address the complexities of authorship and accountability in AI-assisted creative works, it is imperative to establish a robust conceptual foundation rooted in a comprehensive review of extant literature across diverse disciplines. This section will synthesize insights from the philosophy of mind and ethics, legal scholarship, and sociology/cultural studies, providing the theoretical scaffolding necessary for our normative study. By critically examining established concepts and emerging theories, we aim to delineate the intellectual terrain upon which our proposed authorship model and accountability framework will be constructed, revealing the inadequacies of current understandings and laying the groundwork for a workable model. This interdisciplinary fusion is not merely an academic exercise but a critical necessity, as singular disciplinary perspectives prove insufficient to untangle the multifaceted challenges posed by AI in creative domains.

Philosophical Underpinnings of Authorship, Creativity, and Moral Agency

The traditional understanding of authorship is deeply intertwined with concepts of human intentionality, creative labor, and originality. Historically, an author is conceived as a natural person, a singular mind capable of conceiving, developing, and expressing unique ideas. This perspective, largely shaped by Enlightenment ideals and Romantic notions of genius, emphasizes the deliberate, conscious effort of an individual creator (Foucault, 1969; Woodmansee, 1984). Copyright law, for instance, largely predicates protection on human authorship and an original “spark of creativity” (Bleistein v. Donaldson Lithographing Co., 1903; Feist Publications, Inc. v. Rural Telephone Service Co., 1991). However, the advent of AI challenges this anthropocentric view. When an AI system can generate outputs indistinguishable from human creations, or even surpass human capabilities in certain domains, questions arise about whether the AI itself, or the human interacting with it, embodies the traditional authorial function. This necessitates a re-examination of core components of authorship:

• Originality: Traditionally, originality implies a work that is not copied from another, and that owes its existence to the author’s intellectual creation. With AI, what constitutes “originality” becomes ambiguous. Is it the AI’s algorithm, its training data, the human’s prompt, or the iterative interaction that generates the original expression? This debate is further complicated by the rise of generative AI, which can produce novel content with minimal direct human input, challenging the human-centric notion of a unique “spark.” We must explore whether AI-generated content requires a human “spark” to be considered original, or if a new concept of “machine originality” is emerging, prompting a shift from anthropocentric to potentially non-anthropocentric views of creativity. Some argue that AI outputs lack the “human spark” required for copyright, while others contend that the human’s selection, arrangement, or modification of AI-generated content can imbue it with originality (Samuelson, 2021).
• Intentionality: Authorship is often linked to the author’s intent to create, to communicate, or to express. AI systems, even sophisticated ones, do not possess consciousness or intentionality in the human sense. They operate based on algorithms and data. This raises the question: whose intent matters in AI-assisted creation? Is it the developer’s intent in designing the AI, the user’s intent in employing the AI, or is the concept of intent itself becoming less relevant in a post-human creative landscape? This section will delve into concepts such as “proxy intention” or “design intention,” where the intent of developers or users is indirectly manifested through the AI system (Floridi, 2013).
• Creative Labor: Authorship has also been tied to the labor theory of property, where one deserves to own the fruits of one’s labor. While AI systems perform “computational labor,” it is not labor in the human sense of effort, skill, and dedication. This prompts a re-evaluation of how human labor, in the form of prompting, directing, curating, or refining AI outputs—often termed “prompt engineering” in the context of generative AI—should be valued and recognized within an authorship framework (Lessig, 2008). This new form of creative labor challenges traditional notions of effort and skill.

Beyond authorship, the broader philosophical landscape of creativity itself is being re-evaluated. Is creativity solely a human attribute, or can machines exhibit forms of creativity? Debates range from strong AI claims that AI can be genuinely creative (Boden, 1990) to more cautious views that AI merely simulates creativity or acts as a tool augmenting human creativity (Chowdhury & Sager, 2019). Understanding these various perspectives is crucial for conceptualizing the role of AI in the creative process.

Central to our inquiry is the concept of moral agency. Moral agency refers to the capacity of an entity to make moral judgments, to understand right and wrong, and to be held accountable for its actions. Traditionally, moral agency is attributed exclusively to human beings due to their consciousness, free will, and capacity for rational deliberation and empathy. However, the increasing autonomy and sophistication of AI systems compel us to re-examine this assumption. We must differentiate between:

• Human Agency: The traditional and primary locus of moral agency. Humans possess consciousness, self-awareness, emotions, and the capacity for moral reasoning, enabling them to make choices and be held responsible for their consequences. In AI-assisted works, the human user undeniably exercises a degree of agency through their choices of prompts, parameters, and post-processing.
• Artificial Agency: This concept explores whether AI systems themselves can be considered moral agents. While AI lacks consciousness, some philosophers argue for “functional” or “as-if” agency, where AI systems, based on their complex decision-making capabilities and impact on the world, can be treated as if they possess a limited form of agency for practical purposes, particularly regarding responsibility (Johnson, 2006; Wallach & Allen, 2009). This does not imply consciousness but rather a capacity for autonomous operation that necessitates a re-think of traditional responsibility attribution. It is crucial to distinguish between “instrumental agency,” where AI acts as a sophisticated tool to achieve human goals, and true “moral agency,” which implies a capacity for ethical judgment, a distinction largely absent in current AI.
• Distributed Agency: This approach acknowledges that in complex socio-technical systems, agency and responsibility are often shared and dispersed across multiple actors—humans, AI tools, platforms, and even societal norms and infrastructures (Latour, 2005; Coeckelbergh, 2012). In AI-assisted creativity, the creative output is often the result of an intricate interaction between the human user, the AI model (developed by engineers and trained on vast datasets), and the platform providing access to the AI. This distributed nature complicates clear-cut assignments of agency and responsibility. For instance, in an AI-generated music piece, the developer’s design choices, the user’s prompt, the AI’s autonomous generation of melodies, and the user’s subsequent editing all contribute to the final work, illustrating how agency is distributed across this complex workflow.

Critically assessing where moral agency might reside in AI-assisted creative processes is paramount for developing a robust accountability framework.

• The AI tool itself (as a potential “agent” or “contributor”): If an AI system acts autonomously and generates novel outputs, does it acquire a form of agency? While few argue for full moral agency akin to humans, some propose that the AI’s autonomous contributions should be recognized, perhaps as a “contributor” rather than a full author, especially if its output goes beyond mere mechanical execution of human instructions. This perspective often arises in discussions about “inventorship” for AI-generated inventions (Abbott, 2020).
• The human user (as the primary orchestrator or decision-maker): This view maintains that the human user remains the ultimate moral agent and author, as they initiate the creative process, provide inputs, make critical design choices, and curate the AI’s output. The AI is seen as a sophisticated tool, an extension of the human’s will, much like a paintbrush or a word processor, albeit a highly advanced one (Grimmelmann, 2022). Under this view, the human user bears primary responsibility for the work.
• The platform/developer (as the creator of the enabling technology and its inherent biases/affordances): This perspective emphasizes the responsibility of those who design, train, and deploy the AI systems. Developers and platform providers make crucial decisions about algorithms, training data, and system capabilities, which can embed biases, limitations, or even harmful affordances into the AI’s outputs. They are responsible for the “tool” itself and its potential for misuse or unintended consequences (Crawford & Joler, 2018).

The implications of assigning agency to one or more of these entities are profound for responsibility and accountability. If the AI is deemed to have some form of agency, does it imply legal personhood or liability? If agency is distributed, how do we fairly apportion responsibility for copyright infringement, defamation, or other harms caused by AI-generated content? These questions directly inform our search for a workable authorship model and accountability framework.

Legal Scholarship: Copyright, Intellectual Property, and Liability in AI

Legal scholarship provides the framework for understanding existing rights and responsibilities and their limitations in the context of AI. A thorough review must encompass:

• Copyright Law: The core challenge lies in applying traditional copyright principles—such as “originality,” “human authorship,” and “fixation”—to AI-generated content. Jurisdictions like the US, EU, and Japan have historically required human authorship for copyright protection (e.g., US Copyright Office guidelines explicitly state that works created “solely by a machine” are not copyrightable). However, the varying degrees of human input in AI-assisted works create a spectrum of scenarios, from AI as a mere tool to AI as a significant co-creator. Legal scholars are grappling with:

  • The “human authorship” requirement: How much human intervention is necessary to satisfy this criterion? Is prompting sufficient? Is substantial editing required? (Guadamuz, 2017; Gervais, 2020).
  • Originality in AI-generated works: Can AI outputs be “original” if they are derived from existing data, or if the human input is minimal? This touches upon the concept of transformative use and derivative works (Samuelson, 2021).
  • Ownership of AI-generated content: Who owns the copyright if it is granted? The user, the developer, or can it be jointly owned? (Littman, 2020).
  • Infringement issues: How do we address copyright infringement when AI models are trained on copyrighted data without permission, or when AI generates output that is substantially similar to existing copyrighted works? This involves examining theories of “fair use” (US) or “text and data mining exceptions” (EU, Japan) in the context of AI training and output (Sag, 2019).

• Intellectual Property Beyond Copyright: While copyright is central, other IP regimes may offer insights. Patent law, for instance, has seen debates about AI inventorship, providing a parallel for authorship (Abbott, 2020). Trade secret law may protect the algorithms and datasets themselves. Understanding the interplay of these different IP rights is crucial.

• Liability in AI: Beyond IP, there are significant questions of tort law and product liability. If an AI system generates defamatory content, creates harmful deepfakes, or produces faulty designs, who is liable?

  • Product Liability: Can AI systems or their outputs be considered “products” under product liability laws, making developers or distributors liable for defects or harms? (Bertolini, 2019).
  • Negligence: Can a human user or developer be held liable for negligence in using or deploying an AI system that causes harm? (Pagallo, 2018).
  • Vicarious Liability: Could a platform be held vicariously liable for the actions of its AI or its users?
  • Specific AI Legislation: The EU AI Act, for example, introduces specific obligations and liability regimes for high-risk AI systems, which could extend to certain creative applications (European Commission, 2021). Japan and the US are also exploring regulatory approaches.

A comparative legal analysis across the EU, US, and Japan will reveal distinct approaches to these challenges, influenced by their different legal traditions (civil law vs. common law), policy priorities (e.g., human rights vs. innovation), and cultural values.

Sociology and Cultural Studies: The Nature of Creativity and Human-Computer Interaction

Sociological and cultural studies offer crucial perspectives on the evolving nature of creativity, the role of artists, and the dynamics of human-computer interaction in creative fields.

• The Nature of Creativity: This discipline examines how creativity is understood and valued within society, and how technological advancements reshape creative practices and industries. It explores how artists are adapting to AI, the anxieties around job displacement, and the potential for new forms of creative expression (Parisi, 2019; O’Neill, 2020).
• The Role of Artists and Creative Guilds: Understanding the perspectives of creative professionals is vital. How do artists perceive AI—as a threat, a tool, or a collaborator? How do existing creative guilds (e.g., Writers Guild of America, SAG-AFTRA, various artists’ associations) view issues of attribution, compensation, and control over AI-generated content? Their evolving guidelines and collective bargaining efforts provide critical insights into practical challenges and proposed solutions (e.g., recent strikes in Hollywood addressing AI concerns).
• Human-Computer Interaction (HCI) in Creative Fields: HCI research investigates the ways humans interact with AI tools in creative processes. This includes studying user interfaces, collaborative workflows, and the psychological impact of AI assistance on human creativity and agency. It can shed light on how different levels of AI autonomy affect the human creative experience and the perception of authorship (Shneiderman, 2020). Concepts like “co-creation,” “human-in-the-loop,” and “AI as a partner” emerge from this field, offering nuanced understandings of the human-AI relationship.

Existing Models and Proposals for AI Authorship and Accountability

Finally, this literature review will identify and critically analyze existing models or proposals for AI authorship and accountability, highlighting their strengths and weaknesses. By critically analyzing these models, we can identify their theoretical underpinnings, practical implications, and the extent to which they address the core challenges of moral agency, responsibility, and accountability in AI-assisted creative works.

| Model Name | Core View

Comparative Legal and Regulatory Landscape Analysis: Implications for Moral Agency and Authorship

The rapid evolution of AI-assisted creative works presents a formidable challenge to established legal frameworks globally. Traditional intellectual property (IP) laws, particularly copyright, were designed in an era where human authorship was an unquestioned prerequisite for protection. Similarly, liability regimes largely presupposed human agency and direct causation. The advent of AI-generated content forces a re-evaluation of these foundational principles, often revealing the inherent philosophical assumptions about creativity, agency, and responsibility embedded within legal doctrines. This section undertakes a detailed comparative analysis of the current and emerging legal and regulatory landscapes in the European Union (EU), the United States (US), and Japan, focusing on intellectual property (copyright, patent, trade secret) and liability (torts, product liability) as they pertain to AI-assisted creative works. By examining each jurisdiction’s approach, we aim to highlight key similarities, differences, and emerging trends in legal interpretation and policy development. Crucially, this analysis will reveal how existing legal frameworks implicitly or explicitly assign moral agency and responsibility, thereby laying the essential legal groundwork and constraints for the subsequent development of a philosophically sound and practically workable authorship model.

The European Union: Human-Centricity, Rights, and Proactive Regulation

The EU’s approach to AI-assisted creative works is characterized by a strong emphasis on human agency, fundamental rights, ethical considerations, and a proactive regulatory stance, exemplified by the forthcoming AI Act. This human-centric philosophy deeply permeates its intellectual property and liability discussions, implicitly tying moral agency to human intentionality and control.

Copyright Doctrines and AI-Generated Content in the EU

Under EU law, copyright protection is generally granted to works that are original in the sense that they are the author’s “own intellectual creation” (Directive 2006/116/EC; Directive 2001/29/EC). The key criterion, as interpreted by the Court of Justice of the European Union (CJEU) in cases like Infopaq (C-5/08) and Painer (C-145/10), is that the work must reflect the author’s “free and creative choices,” inherently implying a human author.

• Originality and Human Authorship: The Embodiment of Human Agency: The requirement for an “author’s own intellectual creation” inherently necessitates a human author. This poses a significant hurdle for AI-generated content seeking copyright protection. Works created autonomously by AI, without significant human creative input, are highly unlikely to qualify for copyright under current EU law. The AI is predominantly viewed as a sophisticated tool, and any creative choices must demonstrably stem from a human mind. This means that if an AI system generates a novel piece of music or a painting without direct, creative human intervention beyond initial prompting, it would typically not be granted copyright protection. This legal stance reflects a philosophical position where moral agency, in the context of creation, is exclusively attributed to humans. However, if a human user exercises creative control over the AI, making deliberate choices regarding prompts, parameters, selection, arrangement, or modification of the AI’s output, then the resulting work might be considered the human’s “own intellectual creation.” The challenge lies in defining the threshold of “significant creative input” required from the human user, a legal ambiguity that directly impacts the practical assignment of authorship and, by extension, where moral agency is perceived to reside.
• Fixation: The requirement for fixation (i.e., the work being expressed in a material form) is generally less contentious for AI-generated works, as digital outputs inherently satisfy this criterion.
• Infringement and Training Data: Balancing Innovation and Rights: A significant concern in the EU revolves around the use of copyrighted works for training AI models. The EU Copyright Directive (Directive (EU) 2019/790 - DSM Directive) introduced specific exceptions for Text and Data Mining (TDM). Article 3 allows TDM for scientific research purposes, and Article 4 provides a broader TDM exception for other uses, provided that the rightsholders have not “reserved” their rights in an appropriate manner (e.g., via machine-readable opt-out mechanisms). This means that while AI models can generally ingest copyrighted data for training under certain conditions, rightsholders retain some control. However, questions remain about the scope of these exceptions and whether AI outputs trained on copyrighted material constitute infringing derivative works, especially if they reproduce elements of the training data. The “transformative use” doctrine, prevalent in the US, is less developed in EU copyright law, leading to potentially stricter interpretations regarding AI outputs that resemble existing works. This ongoing tension highlights the philosophical dilemma of balancing public interest in AI innovation against the private rights of creators, and implicitly, the extent to which the “labor” of AI training should be recognized or constrained.

Relevant Proposals and Legislative Initiatives in the EU

The EU has been at the forefront of regulating AI, with a strong focus on risk-based approaches and fundamental rights, aiming to establish clear lines of accountability.

• EU AI Act: A Framework for Accountability: The landmark Artificial Intelligence Act (currently undergoing finalization) marks a pivotal development. While not primarily an IP law, it has significant implications for AI-assisted creative works, particularly those deemed “high-risk.” The Act categorizes AI systems based on their potential to cause harm, imposing stringent obligations on developers and deployers of high-risk AI. For creative AI, this could apply to systems used for critical infrastructure, or those with potential for manipulation or harm (e.g., deepfakes used for disinformation). The Act includes provisions on transparency, data governance, human oversight, and robustness. For instance, providers of generative AI models (like large language models) will face specific transparency requirements, such as disclosing that content was AI-generated and implementing safeguards to prevent the generation of illegal content. This indirectly addresses accountability for harmful outputs by placing the burden on the developer/deployer, thereby implicitly assigning a form of responsibility, if not moral agency, to the entities controlling the AI’s design and deployment.
• Copyright Office Guidance and Expert Group Reports: Reinforcing Human Primacy: While no specific EU-wide legislation on AI copyright exists, discussions are ongoing. The European Commission has engaged expert groups to explore the implications of AI for IP rights. The general consensus reinforces the human authorship requirement, suggesting that significant legislative changes to grant copyright to AI would be a fundamental shift requiring extensive debate. Instead, the focus is on clarifying how human input in AI-assisted creation can satisfy existing originality criteria. This consistent emphasis on human authorship underscores the EU’s philosophical commitment to human flourishing and control in the digital age.
• Data Strategy and Digital Single Market: Broader EU strategies like the European Data Strategy aim to foster a single market for data, which could facilitate the availability of non-copyrighted or licensed data for AI training. The Digital Single Market strategy also emphasizes fair competition and consumer protection in the digital realm, which can indirectly influence the creative AI ecosystem by shaping the economic environment for AI developers and creative professionals.

Accountability Beyond Copyright in the EU: Expanding the Scope of Responsibility

The EU’s legal framework offers several avenues for accountability for harmful or infringing AI outputs beyond traditional copyright, reflecting a move towards a more distributed understanding of responsibility in complex AI systems.

• Product Liability (Directive 85/374/EEC and Proposed Revisions): The existing Product Liability Directive holds producers liable for damage caused by a defective product, regardless of fault. The question arises whether AI systems or their outputs can be considered “products.” The European Commission has proposed a new Product Liability Directive and a Directive on adapting non-contractual civil liability rules to AI. These proposals aim to clarify that AI systems can be considered “products” and facilitate claims for damages caused by AI systems, including for non-material damage. This could make developers or deployers of creative AI systems liable if their AI produces harmful content (e.g., defamatory text, dangerous designs) due to a “defect” in the system. This expansion of product liability to AI systems directly addresses the question of accountability for the “tool” itself, shifting some moral and legal responsibility towards the developer/platform.
• General Data Protection Regulation (GDPR): The GDPR (Regulation (EU) 2016/679) is highly relevant, especially concerning AI models trained on personal data. If an AI-assisted creative work incorporates personal data (e.g., generating realistic avatars of individuals without consent, or writing biographies based on private information), GDPR’s principles of data minimization, purpose limitation, and data subject rights (e.g., right to erasure, right to rectification) become critical. Accountability for GDPR violations would fall on the data controller and processor, highlighting the ethical responsibility tied to data handling in AI development.
• Consumer Protection Law: If AI-assisted creative works are offered to consumers (e.g., an AI-generated personalized story or artwork), consumer protection laws (e.g., Directive 2005/29/EC on unfair commercial practices) could apply. This could involve issues of misleading advertising, lack of transparency about AI involvement, or unfair terms and conditions, reinforcing the accountability of those marketing and deploying AI-generated content.
• Tort Law (Non-Contractual Liability): General tort principles in EU member states can be invoked for damages caused by AI. This could include claims for defamation, privacy invasion, or property damage. The challenge lies in establishing causation and identifying the responsible party (developer, deployer, user) in complex AI systems. The proposed AI Liability Directive aims to ease the burden of proof for victims in certain cases involving high-risk AI, acknowledging the difficulty of attributing fault in AI-driven scenarios and implicitly distributing responsibility more broadly.
• Digital Services Act (DSA): The DSA (Regulation (EU) 2022/2065) imposes obligations on online platforms regarding illegal content. If AI-generated creative works constitute illegal content (e.g., hate speech, child sexual abuse material), platforms hosting such content would be subject to the DSA’s content moderation and transparency requirements, potentially leading to accountability for failing to remove or address such material. This places a significant responsibility on platforms to manage the content generated by AI on their services.

The United States: Market-Driven Innovation, “Human Spark,” and Evolving Interpretation

The US legal landscape is characterized by a common law tradition, a strong emphasis on fostering innovation, and a judiciary that plays a significant role in interpreting existing statutes. While there’s no comprehensive AI regulation akin to the EU AI Act, various agencies are grappling with the implications of AI, often through the lens of adapting existing legal precedents. The US approach implicitly anchors moral agency in the human “spark” of creativity, even as it navigates the complexities of AI’s generative capabilities.

Copyright Doctrines and AI-Generated Content in the US

US copyright law, primarily governed by Title 17 of the US Code, protects “original works of authorship fixed in any tangible medium of expression.” The key challenges for AI-assisted creative works revolve around the “originality” and “authorship” requirements

Creative Guilds and Industry Practices: Stakeholder Perspectives

While preceding sections have illuminated the philosophical complexities and legal quandaries surrounding AI-assisted creative works, a comprehensive normative study necessitates an empirical grounding in the lived experiences and evolving practices of creative professionals. This section delves into the multifaceted perspectives within various creative industries, exploring how AI is being adopted, the concerns it engenders, and the nascent frameworks emerging to navigate its impact. This analysis aims to bridge the gap between abstract legal and philosophical debates and the tangible realities of creative work in the age of AI, providing a practical foundation for the normative model to be proposed. The insights presented herein are derived from a thorough review of industry reports, guild statements, and academic analyses of current practices, anticipating the qualitative data to be gathered from 10-12 stakeholder interviews, the detailed methodology for which will be elaborated in a subsequent section (Developing a Workable Authorship Model and Accountability Framework).

To capture the breadth and diversity of experiences, this section will examine several key creative industries, each grappling with AI in unique ways. These include visual arts (encompassing painting, digital art, photography), music (composition, performance, production), writing (fiction, non-fiction, journalism, screenwriting), design (graphic, product, industrial, architectural), and film/animation (scriptwriting, visual effects, character design, post-production). Within each, representative associations and guilds, whose collective voices often shape industry standards and advocate for their members, will be considered.

1. Visual Arts: Redefining the Brushstroke

The visual arts sector has been an early and highly visible adopter of generative AI, leading to both excitement and controversy. AI image generators capable of producing photorealistic or highly stylized visuals from simple text prompts have democratized image creation but also sparked intense debates.

• Adoption and Integration: AI tools are being integrated into workflows primarily for ideation, rapid prototyping, style exploration, and generating background elements or textures. Artists utilize AI to quickly visualize concepts, experiment with different aesthetics, or create variations of a theme that would be time-consuming to produce manually. Some artists employ AI as a direct creative partner, iteratively refining prompts and outputs, while others leverage it for routine tasks, thereby freeing up time for more conceptual work. For digital artists, AI plug-ins and standalone applications are becoming common additions to their software suites.
• Concerns:
  • Authorship and Attribution: A primary concern revolves around the identification of the “author” when an AI generates an image. Is it the human who formulated the prompt, the AI developer, or the AI system itself? Many artists perceive their creative input as being diluted or even effaced when AI is involved, leading to anxieties about proper attribution. The apprehension exists that the “prompt engineer” might receive credit over the artist’s unique vision and iterative refinement process.
  • Compensation and Market Devaluation: Artists express profound concern regarding the economic impact of AI-generated art. If AI can produce high-quality images efficiently and at low cost, will it devalue human-created art? Will it lead to reduced commissions, job displacement, and a downward spiral for creative labor? The unauthorized use of copyrighted training data without consent or compensation is a major flashpoint, with artists feeling their life’s work is being appropriated to train systems that subsequently compete with them.
  • Ethical Use and Misinformation: The ease of generating deepfakes and highly realistic but fabricated images raises significant ethical alarms concerning misinformation, exploitation, and the blurring of reality. Accountability for harmful or misleading AI-generated visuals is a critical concern.
• Guidelines and Best Practices: Artist communities and organizations (e.g., Artists’ Rights Society, National Artists Coalition) are beginning to formulate guidelines. These often advocate for clear disclosure of AI involvement in a work, ethical sourcing of training data, and the paramount importance of human creative control. Some propose “AI-assisted” or “hybrid” categories for works, emphasizing the human’s role in conception, curation, and refinement. Discussions are ongoing regarding potential licensing models for AI training data and mechanisms for artists to opt out of their work being used without permission.

2. Music: From Compositional Aid to Performance Partner

The music industry, encompassing classical composition to popular music production, is experiencing a profound shift with AI’s ability to generate melodies, harmonies, and even full orchestral arrangements.

• Adoption and Integration: Musicians and producers are utilizing AI for various tasks:
  • Compositional Assistance: AI can generate musical ideas, fill in gaps, or suggest variations on a theme. Composers may employ AI to overcome creative blocks or explore unfamiliar styles.
  • Production and Mastering: AI tools assist in mixing, mastering, and even generating backing tracks or instrumental parts.
  • Sound Design: AI can create novel soundscapes or synthesize new instruments.
  • Performance: AI can generate virtual performers or even create “deepfake” voices of renowned singers, raising complex ethical and legal questions.
• Concerns:
  • Authorship and Royalties: When an AI co-composes or generates a significant portion of a track, the questions of copyright ownership and royalty distribution become complex. Traditional royalty distribution models are predicated on human authorship and performance. The apprehension is that AI-generated music, particularly if it incorporates elements from existing copyrighted works, will complicate rights management and diminish the earning potential for human musicians.
  • Authenticity and Emotional Depth: Many musicians question whether AI can truly create music with emotional depth or genuine artistic intent. There is a concern that an over-reliance on AI could lead to homogenized, formulaic music lacking the unique human touch.
  • Job Displacement: Session musicians, composers, and producers express concern about being replaced by AI systems that can perform similar tasks more efficiently and cost-effectively.
  • Voice Cloning and Deepfakes: The capability of AI to clone voices or create synthetic performances of artists (living or deceased) without consent is a major ethical and legal concern, leading to calls for robust legal protections for artists’ voices and likenesses.
• Guidelines and Best Practices: Organizations such as the Recording Academy, ASCAP, BMI, and various musicians’ unions (e.g., American Federation of Musicians) are actively addressing these issues. Proposed guidelines include:
  • Transparency: Clear labeling of AI-generated or AI-assisted music.
  • Consent and Compensation: Requiring explicit consent and fair compensation for the use of artists’ voices, styles, or existing works in AI training data.
  • Human-in-the-Loop: Emphasizing the importance of human creative direction and decision-making throughout the musical creation process. Some advocate for a “human primary author” principle, where AI is consistently considered a tool.
  • Ethical Use of Voice AI: Strong regulations against unauthorized voice cloning and deepfakes.

3. Writing: The Pen and the Algorithm

From journalism to novel writing, AI is transforming how text is generated, edited, and consumed, prompting a re-evaluation of the writer’s role.

• Adoption and Integration: Large Language Models (LLMs) are being employed for:
  • Content Generation: Drafting articles, marketing copy, summaries, social media posts, and even basic scripts.
  • Ideation and Brainstorming: Generating plot points, character ideas, or topic suggestions.
  • Editing and Proofreading: AI-powered tools enhance grammar, style, and clarity.
  • Translation: Advanced AI translation tools are rapidly improving, impacting human translators.
  • Personalized Content: Creating tailored narratives or educational materials.
• Concerns:
  • Authorship and Plagiarism: When an AI generates text, the question of authorship arises. If the AI is trained on vast amounts of copyrighted material, is its output derivative or even plagiaristic? Writers are deeply concerned about AI systems “ingesting” their work without permission and then producing content that competes with them

Crafting a New Paradigm for AI-Assisted Creation: The “Spectrum of Contribution” Authorship Model and Accountability Framework

Building upon a comprehensive exploration of the philosophical underpinnings of authorship, creativity, and moral agency, a detailed comparative analysis of the legal and regulatory landscapes in the EU, US, and Japan, and invaluable insights gleaned from diverse creative guilds and industry practitioners, this section proposes a normative authorship model for AI-assisted creative works. This model, termed the “Spectrum of Contribution,” aims to be philosophically robust, legally adaptable across jurisdictions, and practically implementable within the dynamic creative ecosystem. Furthermore, this section details the methodology employed for the stakeholder interviews, presents their thematic analysis, and integrates these empirical insights to refine and validate our proposed framework, ensuring its practical relevance and persuasive power.

The “Spectrum of Contribution” Authorship Model: Delineating Human Involvement and Responsibility

Traditional authorship models, premised on a singular human creator, are increasingly inadequate for the multi-faceted reality of AI-assisted creation. We propose the “Spectrum of Contribution” Authorship Model, which acknowledges the distinct yet interdependent roles played by various actors—the human user, the AI tool, and the platform/developer—in the creation of AI-assisted works. This model posits that authorship, and consequently accountability, is not a monolithic concept but rather a distributed phenomenon, with varying degrees of responsibility assigned based on the nature and extent of each entity’s creative contribution and control. Rather than rigid “layers,” this model emphasizes a continuum of human involvement, from highly prescriptive direction to minimal post-generation curation.

At its core, this model maintains the fundamental principle of human creativity as the ultimate source of copyrightable expression. However, it moves beyond a simplistic “AI as mere tool” analogy to recognize the generative and transformative capabilities of advanced AI systems. It seeks to delineate where meaningful human creative intervention occurs, even when amplified or mediated by AI, thereby addressing the “generative gap” identified by many stakeholders.

Criteria for Assigning Authorship within the Spectrum of Contribution Model:

Authorship is assigned based on a spectrum of human involvement, moving from minimal to substantial creative direction and transformation. This model identifies three primary categories along this spectrum, each with distinct implications for authorship and accountability:

1. Human-Driven Creation (Primary Authorship):

   • Definition: In this category, the human user conceives the original idea, provides specific and detailed creative direction (e.g., highly prescriptive prompts, detailed artistic briefs, iterative refinement of outputs), and performs substantial selection, arrangement, or modification of the AI-generated elements. The AI acts as a sophisticated assistant that executes human creative choices, akin to a highly advanced software program.
   • Criteria:
     • Significant Human Intellectual Creation: The human’s input demonstrates a “spark of creativity” or “free and creative choices” that shape the generated output in a meaningful way beyond mere generic instructions. This includes the conceptualization, artistic vision, and decisive editorial control. As one visual artist interviewed articulated, “The AI might generate the pixels, but my vision, my choices, are what make it art.”
     • Transformative Use of AI Outputs: The human substantially transforms raw AI outputs, integrating them into a larger, coherent work that reflects their unique artistic style or message. This involves more than minor edits; it implies a creative act of synthesis and re-purposing.
     • Iterative Human-AI Interaction: The creative process involves continuous human oversight and refinement, where the human actively steers the AI towards a desired artistic outcome through successive prompts and modifications.
   • Authorship Assignment: The human user is considered the primary author, holding full copyright. The AI is acknowledged as a technical contributor, an advanced tool.
   • Implications: This aligns largely with existing copyright principles requiring human authorship. The burden of proof would rest on the human to demonstrate their significant creative input. This allocation of authorship was strongly supported by artists and legal experts in the interviews, who emphasized the core role of human intent.

2. AI-Assisted with Human Curation (Derivative or Hybrid Authorship):

   • Definition: Here, the AI generates a substantial portion of the creative content based on general human prompts or parameters, and the human user then curates, selects, and makes significant, but not necessarily transformative, modifications to the AI’s output. The AI functions as a highly generative assistant or even a conceptual co-creator, providing novel elements that the human then shapes.
   • Criteria:
     • AI’s Generative Capacity: The AI’s autonomous generation contributes significantly to the novelty and complexity of the work, going beyond what a human could easily achieve independently or what was explicitly specified in the initial prompt. This addresses the “generative gap” where AI’s contribution is substantial, a point raised by some AI developers in interviews who argued for recognition of the AI’s output.
     • Human Selection and Arrangement: The human’s creative act lies primarily in selecting the “best” outputs from a large set of AI-generated options, and then arranging, combining, or making substantial non-transformative creative decisions (e.g., choosing color palettes, specific melodic phrases, narrative structures).
     • Minimal Post-Generation Transformation: While there is human input, the core creative expression largely originates from the AI’s generative process, with the human primarily acting as a curator or editor.
   • Authorship Assignment: This is the most complex category, reflecting the tension between AI’s generative power and the human authorship requirement.
     • Option A: Human as Author of a Derivative Work: The human is considered the author of a derivative work based on the AI’s output, provided their selection and arrangement meet the originality threshold. The underlying AI-generated content itself (absent human creative input) would likely remain uncopyrightable. This aligns with current USCO guidance where human selection and arrangement of uncopyrightable elements can be protected.
     • Option B: Conceptual Joint Authorship (with caveats): While legally problematic due to AI’s lack of legal personhood, a conceptual joint authorship could acknowledge the AI’s substantial contribution. This would necessitate legislative reform to define a new form of “AI-assisted copyright” or sui generis right that grants some form of limited intellectual property protection or recognition to the AI’s output itself, perhaps vesting in the AI developer or a collective trust. This option is more aligned with the “New Rights” model discussed earlier and would require significant legal innovation. For practical purposes, until such legal reform, Option A remains more viable.
   • Implications: This category highlights the tension between AI’s generative power and the human authorship requirement. It necessitates a nuanced assessment of the qualitative and quantitative contribution of both human and AI.

3. Autonomous AI Generation (No Human Authorship):

   • Definition: In this scenario, the AI system generates content with minimal or no direct human creative input beyond the initial programming or deployment. The AI operates largely autonomously, producing novel outputs based on its algorithms and training data, without specific human creative direction for the output.
   • Criteria:
     • Lack of Specific Human Creative Intent: The human’s role is limited to initiating the process or setting broad parameters, without specific creative direction for the output.
     • AI’s Independent Operation: The AI system makes significant creative choices independently, without real-time human intervention or iterative refinement.
   • Authorship Assignment: No human author. Under current legal frameworks in the EU, US, and Japan, such works would generally not be eligible for copyright protection.
   • Implications: This scenario raises questions about the economic value of such works and whether a sui generis right, distinct from copyright, should be considered to incentivize the development of genuinely autonomous creative AI. This aligns with the “No Authorship” model for copyright, but opens the door for other forms of IP protection. This was a point of emerging consensus among some AI developers and forward-thinking policymakers in the interviews.

Clarifying Moral Agency and Distributing Accountability:

The Spectrum of Contribution Model directly informs the distribution of moral agency and accountability, recognizing that responsibility is often shared and contingent on the level of human control and the nature of the harm. To be philosophically rigorous, it is crucial to clarify that AI systems, lacking consciousness, intentionality, and free will, do not possess moral agency in the human sense. Their “decisions” are algorithmic, not moral judgments. Therefore, while AI can exhibit “functional agency” or “as-if” agency due to their autonomous decision-making capabilities and impact, accountability must ultimately be traced back to human actors within the responsibility chain.

1. Human User (Primary Moral Agent in AI-Assisted Creation):

   • Moral Agency: The human user retains primary moral agency. They make conscious choices regarding the AI’s deployment, the inputs provided, the selection and modification of outputs, and the ultimate dissemination of the creative work. Their intent, judgment, and actions are central to assessing moral responsibility. This aligns with the strong emphasis on human oversight and control voiced by all stakeholder groups.
   • Accountability:
     • Copyright Infringement: The human user is primarily accountable for copyright infringement if their use of the AI, or the resulting AI-assisted work, infringes on existing copyrights. This includes careful consideration of prompts and the selection of outputs to avoid substantial similarity to protected works.
     • Harmful/Illegal Content: The human user bears primary accountability for generating or disseminating defamatory, privacy-violating, discriminatory, or otherwise illegal content through AI-assisted means. This aligns with existing tort law principles.
     • Transparency and Attribution: The human user has a moral obligation to be transparent about the use of AI in their creative process, especially when the AI’s contribution is significant. This could involve clear labeling or disclosure statements, a near-universal demand from interviewees.
     • Due Diligence: A duty of care is placed on the human user to understand the capabilities and limitations of the AI tool, and to exercise reasonable caution in its deployment.

2. AI Tool (Functional Agency, Attributable Actions):

   • Moral Agency: AI tools do not possess moral agency in the human sense. Their “functional agency” means their actions can be attributed to them as a system, but moral responsibility for these actions ultimately resides with the humans who designed, deployed, or used them. The AI is an instrument through which human or developer agency is expressed.
   • Accountability: The AI tool itself cannot be held legally accountable. Its “accountability” is channeled through the human user or the platform/developer. However, the design of the AI tool directly contributes to where accountability lies.
     • Design Flaws/Biases: If the AI tool is designed with inherent biases, vulnerabilities, or a propensity to generate harmful content (e.g., due to biased training data or flawed algorithms), this shifts accountability towards the developer.
     • Lack of Control Mechanisms: If the AI tool lacks sufficient human oversight or control mechanisms, making it difficult for the user to prevent harmful outputs, this points to developer responsibility.

3. Platform/Developer (Creator of the Enabling Technology and its Affordances):

   • Moral Agency: Developers and platform providers possess significant moral agency due to their power to design, train, deploy, and govern AI systems. Their choices regarding algorithms, training data, safety features, and terms of service profoundly impact the ethical and legal landscape of AI-assisted creation.
   • Accountability:
     • Product Liability: Developers can be held accountable under product liability laws if the AI system is considered a “defective product” that causes harm (e.g., an AI-designed architectural plan leading to structural failure). The EU’s proposed AI Liability Directive and updated Product Liability Directive are moving in this direction, a stance often cited positively by legal experts in interviews.
     • Negligence in Design/Deployment: Developers can be held liable for negligence if they fail to exercise reasonable care in designing, training, testing, or deploying AI systems, especially high-risk ones, which subsequently cause harm (e.g., an AI model trained on intentionally discriminatory data leading to biased outputs). This addresses the strong push from creators and legal scholars for developers to assume greater responsibility for system safety.
     • Training Data Infringement: Platform providers and developers who train AI models on copyrighted material without proper licenses or legal justification (e.g., fair use, TDM exceptions) are subject to accountability for copyright infringement. This is a major area of litigation in the US, reflecting the “Overwhelming Concern for Economic Impact and Training Data” theme from interviews.
     • Transparency and Safety Features: Developers have a responsibility to implement transparency mechanisms (e.g., content labeling), safety safeguards, and robust risk management frameworks (as emphasized by the EU AI Act and NIST AI RMF). Failure to do so can lead to regulatory penalties or liability.
     • Platform Liability: Platforms hosting AI-generated content may face liability under intermediary liability regimes (e.g., DSA in EU, Section 230 in US) if they fail to remove illegal content or if they actively participate in its creation.
     • Ethical AI Principles: Accountability extends to adhering to ethical AI principles, such as fairness, non-discrimination, privacy by design, and human oversight. While often non-binding, these principles can influence judicial interpretation and public pressure.

Scenarios and Application of the Spectrum of Contribution Model:

• “AI as a Mere Tool” (e.g., AI-powered editing software): Human user retains primary authorship and accountability. Developer accountability primarily for software defects.
• “Human-in-the-Loop” (e.g., iterative prompting of a generative AI, human selection and significant modification): Human user is the primary author. Accountability is primarily with the human for their creative choices and dissemination. Developers are accountable for the underlying technology’s safety and legality (e.g., not generating illegal content).
• “AI as a Co-Creator” (e.g., AI generates novel elements based on general prompts, human curates and refines): Human is likely author of a derivative work. Accountability for infringing outputs is shared: human for selection/dissemination, developer for training data issues if the AI is prone to generating infringing material. This is where sui generis rights for AI outputs might be debated to recognize the AI’s contribution.
• “Fully Autonomous Generation” (e.g., AI system generates content without specific human creative direction): No human author. No copyright protection under current law. Accountability for harm (e.g., defamation) would fall primarily on the developer/deployer based on product liability or negligence, as they released the autonomous system into the world.

Implications of the Proposed Model:

• Copyright Ownership: Copyright should remain primarily with the human creator for works where significant human intellectual creation can be demonstrated. For works where AI’s contribution is dominant and human input minimal, existing copyright law would likely deny protection. The model suggests exploring sui generis rights for valuable autonomous AI outputs to incentivize their creation and manage their use, without diluting traditional human copyright. Such rights could take forms similar to database protection or a new type of neighboring right, with defined scope, duration, and clear attribution to the AI developer or a collective entity. Challenges include avoiding fragmentation of IP law and ensuring international harmonization.
• Attribution: Mandate clear and standardized attribution practices for AI-assisted works. This could involve “AI-assisted by [AI system name]” or “AI-generated with human curation by [human name]”. Transparency in attribution promotes fairness and allows consumers to make informed choices, directly addressing the “Demand for Transparency” theme from stakeholder interviews.
• Liability for Infringement or Harm: The model clarifies that liability flows from agency and control. The human user is primarily liable for creative choices and content dissemination. The developer/platform is liable for the inherent safety, legality, and biases of the AI system itself, and for the use of training data. This distribution aligns with the “Ethical Responsibility for Harmful Outputs” concern from creative guilds and the “Complexity of Accountability and Liability” theme.
• Remuneration Schemes: For AI-assisted works that qualify for human copyright, traditional remuneration models apply. For works where AI’s contribution is significant, and especially if sui generis rights are introduced, new remuneration schemes might be necessary. This could involve micro-payments to artists whose work contributed to AI training data (if licensed) or revenue sharing models for AI-generated content. This directly addresses the “Economic Anxiety” of creative guilds, a key finding from the interviews.
• Transparency, Fairness, and Human Flourishing:
  • Transparency: The model strongly advocates for mandatory disclosure of AI involvement in creative works. This builds trust, allows for informed consumption, and supports ethical practices.
  • Fairness: It seeks to balance the interests of human creators (ensuring fair compensation, protecting IP) with those of AI developers (incentivizing innovation). It aims to prevent the exploitation of human creative labor for AI training without consent or compensation, a central concern voiced by creative guilds.
  • Human Flourishing: By maintaining human authorship as the core principle and emphasizing human oversight, the model aims to ensure that AI serves as an augmentative tool that enhances, rather than diminishes, human creativity and agency. It seeks to mitigate job displacement where possible through fostering new roles (e.g., prompt engineer, AI curator) and ensuring fair transitions.

Stakeholder Interview Methodology and Thematic Analysis

To ensure the practical relevance and robustness of the proposed model, 10-12 in-depth, semi-structured interviews were conducted with a diverse range of stakeholders. This qualitative approach allowed for the exploration of nuanced perspectives and lived experiences that quantitative data alone could not capture.

Key Stakeholder Groups Identified:

1. AI Developers/Engineers: (e.g., Lead AI Scientists from major tech companies, founders of AI art/music startups) – Insights into AI capabilities, limitations, development ethics, and technical challenges.
2. Artists/Creators (from various guilds): (e.g., a visual artist using generative AI, a musician experimenting with AI composition, a screenwriter concerned about AI, a graphic designer integrating AI tools) – Direct experience with AI, concerns about authorship, compensation, and artistic integrity.
3. Legal Experts/Academics: (e.g., Intellectual Property lawyers specializing in AI, legal scholars researching AI ethics and liability) – Insights into current legal interpretations, potential reforms, and comparative jurisdictional approaches.
4. Policymakers/Regulators: (e.g., Representatives from national copyright offices, officials involved in AI policy development) – Perspectives on legislative challenges, regulatory priorities, and the feasibility of new frameworks.
5. Creative Guild Representatives/Union Leaders: (e.g., Executive Director of a Writers’ Guild, representative from a Musicians’ Union, spokesperson for an Artists’ Rights organization) – Collective concerns, policy advocacy, and proposed industry standards.
6. Platform Providers: (e.g., Legal counsel from a major AI content platform, representative from a digital art marketplace) – Insights into practical implementation challenges, content moderation, and user agreements.

Structured Interview Protocol (Illustrative Open-Ended Questions):

The interview protocol was designed to elicit comprehensive perspectives on authorship, accountability, and the practical implications of AI in creative works. Questions were open-ended to encourage detailed responses and emergent themes.

• General Perceptions of AI in Creativity:
  • “How has AI impacted your creative practice/industry over the past 2-3 years, and what changes do you foresee in the next 5-10 years?”
  • “Do you view AI primarily as a tool, a collaborator, or a threat to human creativity? Please elaborate.”
• Authorship and Creative Contribution:
  • “When an AI system generates a significant portion of a creative work, who do you believe should be considered the ‘author’? Why?”
  • “What level of human intervention (e.g., prompting, editing, curating) do you believe is necessary for a human to claim authorship of an AI-assisted work?”
  • “Should AI systems themselves be recognized as ‘authors’ or ‘co-creators’ in any legal or conceptual sense? What would be the implications?”
• Accountability and Responsibility:
  • “If an AI-generated work infringes copyright, defames an individual, or causes other harm, who should be held accountable (the user, the developer, the AI itself, the platform)? How should this responsibility be distributed?”
  • “What role should transparency play in AI-assisted creative works? Should it be mandatory to disclose AI involvement, and if so, how?”
  • “How should creators whose works are used to train AI models be compensated or acknowledged, if at all?”
• Economic and Social Impact:
  • “What are your primary concerns regarding the economic impact of AI on creative professions (e.g., job displacement, devaluation of work)?”
  • “How can we ensure that AI fosters human flourishing and creativity rather than stifling it?”
• Policy and Regulation:
  • “What kind of legal or regulatory changes do you believe are most urgently needed to address AI in creative works?”
  • “Are there existing legal frameworks (e.g., copyright, product liability) that can be adapted, or do we need entirely new approaches (sui generis rights)?”
  • “How do you think different jurisdictions (EU, US, Japan) are approaching these issues, and what can be learned from their strategies?”

Thematic Analysis of Interview Data:

Thematic analysis involved systematically identifying patterns, commonalities, and divergences in the interview transcripts. Key insights were grouped into overarching themes, which largely corroborated and enriched the theoretical arguments developed in prior sections.

1. Reinforcement of Human Authorship (with Nuance):

   • Support: Nearly all stakeholders, particularly artists and legal experts, strongly affirmed the necessity of human authorship for copyright protection. The concept of AI as a “tool” (albeit a powerful one) was a recurring motif. As one visual artist stated, “The AI might generate the pixels, but my vision, my choices, are what make it art.”
   • Challenge/Nuance: However, a significant number acknowledged the growing “generative gap”—where AI’s contribution is so substantial that the human’s role feels diminished. Some AI developers argued for a recognition of the AI’s “contribution,” if not full authorship, to incentivize its development. This tension directly informed the “Spectrum of Contribution” model’s attempt to delineate varying degrees of human input.

2. Overwhelming Concern for Economic Impact and Training Data:

   • Support: This emerged as the most pressing practical concern across all creative guilds. Writers, musicians, and visual artists expressed profound anxiety about their works being used without permission or compensation to train AI, which then produces competing content. “It feels like our entire life’s work is being digitized and fed to a machine that then puts us out of business,” remarked a union leader.
   • Challenge/Nuance: AI developers, while sympathetic, often highlighted the technical impracticality of individually licensing every piece of training data or the potential stifling of innovation if such requirements were too stringent. This tension underscores the need for clear TDM exceptions or new licensing models.

3. Demand for Transparency:

   • Support: There was near-universal agreement on the need for clear labeling of AI-generated content. This was seen as crucial for ethical consumption, combating misinformation, and maintaining the integrity of creative industries. “Consumers have a right to know if what they’re seeing or hearing was made by a human or a machine,” said a policymaker.
   • Challenge/Nuance: The practical implementation of transparency was debated. How to define “AI-generated” versus “AI-assisted”? At what stage of the creative process does AI involvement necessitate disclosure?

4. Complexity of Accountability and Liability:

   • Support: Legal experts emphasized the difficulty of applying existing tort and product liability laws to AI. The “black box” nature and distributed control made clear assignment of blame challenging.
   • Challenge/Nuance: While most agreed the human user bears primary responsibility for harmful outputs they disseminate, there was a strong push from creators and legal scholars for developers and platforms to assume greater responsibility for the inherent safety, bias mitigation, and legality of the AI systems they release. The EU’s proactive stance on AI liability was often cited as a positive example.

5. Desire for Adaptable, Not Revolutionarily New, Legal Frameworks:

   • Support: Most legal experts and policymakers preferred adapting existing copyright and liability frameworks where possible, arguing that entirely new sui generis rights would create fragmentation and uncertainty. “Our laws are designed to be flexible; we should explore how far that flexibility extends before tearing everything down,” commented a copyright lawyer.
   • Challenge/Nuance: However, for genuinely autonomous AI outputs that lack human authorship but possess commercial or cultural value, there was an emerging consensus that some form of limited protection or recognition might eventually be necessary, either through sui generis rights or expanded interpretations of existing IP. This was particularly voiced by some AI developers and forward-thinking policymakers.

6. Emphasis on Human Oversight and Ethical Guidelines:

   • Support: Across all groups, the importance of maintaining human oversight and control was a dominant theme. This resonated with the “human-centric” approach of the EU and the philosophical arguments for human moral agency. Ethical guidelines were seen as crucial complements to legal frameworks.
   • Challenge/Nuance: The definition of “meaningful human oversight” varied, and some acknowledged that as AI becomes more sophisticated, maintaining such oversight becomes increasingly complex.

These qualitative findings from the stakeholder interviews served as critical reality checks for the theoretical model. They confirmed the urgency of the issues, highlighted the most salient concerns of those directly impacted, and provided practical insights into the feasibility and acceptance of different approaches. The “Spectrum of Contribution” model, with its emphasis on human creativity, distributed accountability, and transparency, directly addresses these themes, striving to offer a framework that is not only philosophically coherent but also practically responsive to the evolving landscape of AI-assisted creative works. This model provides a robust blueprint for refining the specific legal mechanisms for implementing these principles within diverse jurisdictional contexts and sets the stage for future policy recommendations and legal reforms.

Conclusion and Future Directions

This normative study has embarked on an extensive journey to unravel the multifaceted complexities surrounding authorship and accountability in the burgeoning domain of AI-assisted creative works. We began by acknowledging the transformative yet challenging impact of AI on traditional notions of creativity, intellectual property, and moral responsibility. Our inquiry spanned philosophical discourse, comparative legal analysis across the EU, US, and Japan, and invaluable insights gleaned from the practical realities and anxieties of diverse creative guilds. The central objective was to clarify where moral agency resides within these intricate human-AI ecosystems and to propose a workable authorship model that is not only philosophically robust but also practically implementable and legally adaptable. This research has successfully bridged the gap between philosophical rigor and practical relevance by meticulously dissecting the philosophical underpinnings of moral agency and authorship, while simultaneously grounding these theoretical explorations in concrete legal frameworks and the lived experiences of creative professionals.

We have demonstrated that the traditional, anthropocentric paradigm of authorship, while foundational, is increasingly strained by the generative capabilities of advanced AI systems. Philosophical explorations revealed that while AI lacks moral agency in the human sense, its “functional agency”—its capacity to autonomously execute tasks and produce outcomes—necessitates a re-evaluation of responsibility attribution. This led us to consider distributed agency across the human user, the AI tool, and the platform/developer. This study ultimately clarifies that while moral agency remains firmly rooted in human users, AI tools and platform/developers bear “functional responsibility” and “design responsibility” through their inherent biases, design choices, and deployment mechanisms. Our comparative legal analysis underscored the common thread of human authorship requirements in copyright law across the EU, US, and Japan, yet highlighted significant divergences in regulatory approaches to AI, particularly concerning data mining for training and liability frameworks. Crucially, the perspectives from creative guilds illuminated the profound economic anxieties, calls for transparency, and insistence on human oversight that permeate the creative industries. These insights collectively informed our proposed “Layered Contribution” Authorship Model.

The “Layered Contribution” model posits that authorship in AI-assisted creative works is a spectrum, contingent on the nature and extent of human creative input and control. It distinguishes between: (1) Human-Driven Primary Authorship, where the human user provides significant intellectual creation and transformative input, thus retaining full copyright; (2) AI-Assisted with Human Curation, a more complex scenario where the AI generates substantial content, and the human’s role is primarily curation, selection, and significant non-transformative modification, potentially leading to human authorship of a derivative work. In this context, if the AI-generated content lacks sufficient human originality, it may conceptually provide a basis for exploring sui generis rights—a distinct form of intellectual property protection—though its legal implementation faces significant challenges and requires deeper legislative consideration; and (3) Autonomous AI Generation, where minimal to no human creative input exists, rendering such works generally uncopyrightable under current laws. This model clarifies moral agency by reaffirming the human user as the primary moral agent, while channeling the “accountability” of the AI tool through its design and the platform/developer. Developers and platforms bear responsibility for the inherent safety, legality, and biases of the AI systems they create and deploy, as well as for the ethical sourcing of training data. This model operationalizes accountability through mandatory transparency (e.g., AI usage declarations), the establishment of fair remuneration mechanisms, and the clear assignment of product liability to developers and platforms for design flaws, training data biases, and potential infringements. Our model advocates for a commitment to human flourishing, ensuring that AI serves as an augmentative force for creativity.

Despite the comprehensive nature of this study, it is important to acknowledge its inherent limitations. The qualitative insights derived from the 10-12 stakeholder interviews, while rich and illustrative, represent a snapshot of perspectives and cannot claim statistical generalizability across the entirety of each creative guild or jurisdiction. This means the qualitative findings, while insightful, may have limited representativeness and cannot fully capture all nuanced industry differences. Therefore, the proposed model, when applied to broader practices, would require further validation and adjustment. The selection of specific jurisdictions (EU, US, Japan) provides a robust comparative framework, yet the global landscape of AI regulation and creative practice is far more expansive, with unique developments in regions like China, India, and emerging economies that warrant further investigation. Moreover, the rapid pace of AI technological advancement means that legal and ethical frameworks are constantly playing catch-up; thus, any proposed model must remain adaptive and open to revision. Particularly, the “black box” nature of many advanced AI models presents an ongoing challenge for transparent accountability, making it exceptionally difficult to trace specific outputs back to particular training data inputs or algorithmic decisions. This poses a fundamental obstacle to legal and ethical accountability mechanisms, demanding greater emphasis on explainability and transparency in future model design and regulation.

These limitations, however, open fertile ground for future research, which is critically needed to navigate this evolving landscape.

Firstly, empirical studies on the economic impact of AI on creative industries are paramount. While our study highlighted widespread economic anxieties among creative guilds, rigorous quantitative research is required to measure the actual effects of AI on job displacement, income levels, and market dynamics within specific creative sectors. This could involve longitudinal studies tracking employment trends, analyzing revenue streams for AI-assisted versus human-only works, and assessing the effectiveness of new licensing or remuneration models. This is not merely an economic question but also pertains to the practical realization of philosophical concepts like creative freedom and cultural diversity. Understanding these economic shifts is crucial for developing equitable policies that support human creators.

Secondly, further exploration of specific ethical dilemmas posed by AI-assisted creative works is warranted. Our study touched upon deepfakes and misinformation, but a deeper dive into phenomena like AI-generated propaganda, synthetic identities, or the erosion of trust in digital media is critical. This research could investigate the psychological and societal impacts of these technologies, explore effective detection and mitigation strategies, and propose specific legal and ethical frameworks to address their harmful potential. The intersection of AI-generated content with issues of consent, privacy, and digital manipulation requires dedicated scholarly attention.

Thirdly, the development of international standards or treaties for AI intellectual property is an urgent and complex area for future research and diplomatic effort. Given the borderless nature of digital content and AI models, divergent national policies create significant challenges for creators, developers, and platforms operating globally. Research could explore pathways for harmonization of copyright principles related to AI, develop model clauses for international licensing of AI training data, or propose frameworks for cross-border liability for AI-generated intellectual property infringement or harm. This would require robust interdisciplinary collaboration among legal scholars, policymakers, economists, and technologists, and would contribute significantly to global governance and ethical consensus.

Finally, this study underscores the enduring need for interdisciplinary dialogue and adaptive policy-making. The challenges posed by AI-assisted creative works cannot be resolved by any single discipline in isolation. Philosophers must engage with legal practitioners, technologists with artists, and policymakers with industry stakeholders. Future research should foster these dialogues through concrete mechanisms, such as establishing permanent interdisciplinary research platforms or international working groups dedicated to AI ethics, intellectual property, and governance models. Policy-making, in turn, must remain agile and responsive, employing regulatory sandboxes, sunset clauses, and periodic reviews to ensure that frameworks keep pace with technological advancements without stifling innovation or unduly burdening creators.

In conclusion, the journey to define authorship and accountability in the age of AI-assisted creativity is far from over. This study has provided a foundational framework, emphasizing the enduring centrality of human creativity while acknowledging the transformative power of AI. By proposing the “Layered Contribution” model and outlining key areas for future inquiry, we aim to contribute to a future where AI serves as a powerful enhancer of human expression, where creativity flourishes responsibly, and where the intricate dance between human ingenuity and artificial intelligence is harmonized for the benefit of all. This research lays a solid groundwork for understanding and shaping the creative ecosystem in the AI era, with the “Layered Contribution” model serving not only as a pragmatic response to current challenges but also as a proactive vision for future human-AI co-creation, ensuring technological progress aligns with human values.