Towards the Development of a Copyright Risk Checker Tool for Generative Artificial Intelligence Systems
Article No.: 41, Pages 1 - 21
Abstract
Generative Artificial Intelligence (GAI) is fundamentally changing the ways of working and blurring the boundaries between human and machine-generated contents. While there is an increasing interest in the adoption of GAI systems, such as ChatGPT and DALL-E, there are also serious concerns about the copyright of the contents—the inputs or generated as outputs by the GAI systems. Such concerns need to be identified and assessed to ensure the ethical and responsible use of GAI systems. Thus, this article aims to address the key research challenge: “how to identify and assess GAI system's copyright concerns”? In response, we propose the development of a Copyright Risk Checker (CRC) Tool. This tool has been formulated and evaluated using a recognised design science research methodology, drawing on an analysis of 10 legal cases across Australia, the United Kingdom, the United States, and Europe. The CRC Tool has undergone evaluation through an experimental scenario, and the results suggest that it is suitable for conducting an indicative copyright risk check of GAI systems. The outcomes of this preliminary assessment can be further examined by expert legal advisors for an in-depth analysis. The development of the CRC Tool provides a foundation for continued research and advancement in this significant area of study.
1 Introduction
Artificial Intelligence (AI) systems tend to simulate human intelligence in hardware devices and software. They seem to enable them to perform intelligent human oriented tasks, such as learning, reasoning, automated decision making, and problem-solving, with noticeable proficiency [20, 41, 42, 45]. There are several definitions of AI systems. For instance, the Organisation for Economic Cooperation and Development (OECD) defines that “an AI system is a machine-based system that, for explicit or implicit objectives, infers, from the input it receives, how to generate outputs such as predictions, content, recommendations, or decisions that can influence physical or virtual environments. Different AI systems vary in their levels of autonomy and adaptiveness after deployment” [22]. Recently, AI has evolved into Generative Artificial Intelligence (GAI) indicating several potential use cases and benefits across different industries. GAI, a specialised form of AI, presents an impressive ability to generate content (e.g., text, images, reports) that appears creative and original [1].
There is increasing influx of GAI tools such as ChatGPT, DALL-E, Stability AI, and GitHub Copilot that have demonstrated their potential to reshape the way people live, work, learn, and communicate [9, 17, 19]. The impact of GAI's integration into our work and lives is evident from interest in its rapid adoption, as seen when ChatGPT amassed 100 million monthly users within just two months, surpassing the growth rates of popular social media platforms like TikTok and Instagram [16]. Moreover, insights from Fishbowl, an online community for professionals, reveal that a substantial 68% of professionals are already utilising GAI tools, indicating the prevalent influence of this technology [11].
While GAI tools are often applauded for their potential to unlock new possibilities for content generation, automation, and enhancement of human creativity and productivity [7, 14], it is essential to acknowledge the pressing concerns associated, such as the content's copyright infringement, transparency, and trust [10, 43]. This article focuses on copyright concerns, as GAI's inputs and creative outputs blur the lines between human and machine-generated contents and their legal authorship and ownership. This is a significant concern because several companies who created or used GAI systems are now facing legal challenges for copyright infringement, from either the GAI systems’ input/training data or output/generated contents [37]. Consequently, a comprehensive understanding of the copyright concerns surrounding GAI becomes essential to strike a harmonious balance between fostering innovation and creativity while protecting the rights of creators [35, 36]. This article investigates into the intersection of GAI and copyright law, aiming to address the following research question to address the following research goal: to identify and assess the copyright concerns of GAI systems.
This article examines the copyright concerns surrounding GAI by analysing regulations, legislations, and 10 legal cases across Australia, the United Kingdom (UK), the United States (US), and European Union (EU). These specific countries were selected due to the prevalence of GAI copyright cases within their jurisdictions or their active initiatives in developing standards and regulations for AI. The results of this analysis are cast into developing a Copyright Risk Checker (CRC) Tool to identify and assess the GAI concerns. The proposed tool is then evaluated using an experimental scenario. The CRC Tool is a vendor neutral and research-based solution enabling users to identify and assess copyright concerns in their GAI systems. This research is not only crucial for understanding the copyright concerns, but it also serves as a foundation for developing ethically responsible and legally compliant GAI regulations, policies, architecture, and systems.
This article is structured as follows. Firstly, it provides the research background and motivation. Secondly, it discusses the research method. Thirdly, it presents the proposed CRC Tool, and its application. Finally, it discusses the research findings and conclusion with options for future work in this important and emerging area of research.
2 Research Background and Motivation
AI refers to the simulation of human intelligence in machines that are programmed to think and learn like humans [8, 13]. It encompasses various techniques and technologies aimed at enabling machines to perform tasks that typically require human intelligence, such as understanding natural language, recognising patterns, and making decisions [20]. While AI, and more recently GAI systems, seem to offer lucrative benefits [44], they also pose several concerns as well [41, 45]. This section provides research background surrounding current AI regulations, standards, GAI, and its copyright concerns for establishing a strong motivation for this research.
2.1 GAI
GAI is a specialised subset of AI that focuses on generating new content [42], such as images, text, music, and videos [4]. GAI has this ability because of its underpinning models. For instance, Table 1 lists some models that are used to build GAI software systems.
Table 1.
| Models | Definition | References |
|---|---|---|
| Machine Learning | Enables computers to learn from data and improve their performance over time. | [12] |
| Deep Learning | Subset of machine learning that uses neural networks with three or more layers to learn from large data sets. A Large Language Model is a kind of a deep learning model. | [3] |
| Diffusion Processing | Enables computers to analyse and process data in a decentralised manner, simulating how information spreads and influences networks. | [21] |
| Neural Network | Computational learning system that uses a network of functions to understand and translate data inputs into desired outputs. | [12, 18] |
GAI systems model(s) use inputs to generate outputs such as code, audio, images, text, or videos, that resemble human-level intelligence or creativity [2]. The GAI models iteratively learn the patterns and structure from inputs or training data given to the model to generate these outputs that have similar characteristics [4]. The inputs and outputs for a GAI system are visually shown in Figure 1.
Fig. 1.
There are several examples of GAI systems. Some of the key examples of GAI systems with their model type, generative work, and categorisation are noted in Table 2. There is an increasing interest in GAI systems; however, they are also attracting the attention of regulations and standards for ensuring their responsible and safe adoption [40], which are discussed in the next section.
Table 2.
| GAI System | GAI Model | Generated Work | Category of Work by Australian Copyright Definition [48, 49] |
|---|---|---|---|
| ChatGPT, Bard and BingAI | Large Language Model | Language based tasks (question responses, essays, translations, etc.) | Literary works |
| GitHub Copilot | Large Language Model | Programming code | Literary works |
| DALL-E, Midjourney, and Stable Diffusion | Diffusion Model | Images | Artistic work |
| DeepMind and Amazon Polly | Deep Learning | Voice/Speech | Sound recordings |
| MusicLM and Soundraw | Deep Learning | Music | Musical works |
| Stable Diffusion, Make-A-Video and Veed.io | Diffusion Model | Video | Cinematograph films |
2.2 AI Regulations and Standards
This article mainly presents the regulatory AI landscapes in major countries such as Australia, the UK, the US, and the EU, as summarised in Table 3. This is because they present intriguing variations in their approaches, reflecting diverse strategies towards fostering responsible and safe AI development and adoption. These specific countries were selected due to the prevalence of GAI copyright cases within their jurisdictions or their active initiatives in developing standards and regulations for AI. While each country has established ethical frameworks, the lack of binding laws raises concerns regarding the enforceability and uniformity of ethical AI practices globally [1]. The EU's proposed AI Act and the US's non-binding guidelines represent efforts to bridge these gaps [28]. The UK's adaptive, sector-specific strategies, though flexible, risk creating disparities, necessitating stringent oversight for consistency [26, 32]. Australia's collaborative initiatives are commendable, yet they require robust legislative backing for true effectiveness, underscoring the critical need for legal frameworks to accompany ethical guidelines [29, 31]. The recent development around AI management standards, by the International Standardisation Organisation (ISO) and International Electrotechnical Commission's (IEC) Joint Technical Committee (JTC) Subcommittee 42 (ISO/IEC JTC 1/SC 42), may offer significant assistance in this endeavour, providing a standardised framework for AI governance and addressing the existing regulatory gaps.
Table 3.
| Country | Regulation/Standards | Addressed Copyright Concerns | References |
|---|---|---|---|
| Australia | Australia showcases a commitment to ethical AI through initiatives like the Responsible and Inclusive AI whitepaper and the AI Assurance framework, emphasising transparency and accountability. Active participation in global AI standards via ISO/IEC JTC 1/SC 42 further demonstrates Australia's dedication. The impending release of ISO/FDIS 42001 by this committee signifies progress in AI management systems. However, despite these advancements, the absence of specific domestic regulations raises concerns about consistent implementation. While these initiatives provide essential principles, enforceable legislation is vital for translating these ideals into tangible actions. Australia needs comprehensive, binding laws to ensure ethical AI principles are not just guidelines but enforceable standards, fostering a culture of responsible AI practices nationwide. | Transparency, Accountability | [28, 27, 31, 29] |
| United Kingdom (UK) | The UK's sector-based regulatory approach, integrating AI principles into existing frameworks, demonstrates adaptability. Notable legislative efforts like the Data Protection and Digital Information Bill showcase the UK's responsiveness to evolving concerns, particularly regarding data privacy. Financial investments in initiatives such as the Foundation Model Taskforce reflect a commitment to ethical AI deployment. Nevertheless, the sector-specific nature of regulations might create complexities and inconsistencies, making it essential to monitor uniformity in application. | Data Privacy, Adaptability | [26, 33, 32] |
| United States (US) | The United States operates in a decentralised AI regulatory landscape, lacking a nationwide framework. While non-binding guidelines, such as the 'Blueprint for an AI Bill of Rights,' emphasise ethical practices, the absence of binding laws raises questions about the enforceability and standardisation of AI ethics. Local initiatives, while promising, contribute to a fragmented approach, potentially resulting in varying standards and oversight mechanisms. | Ethical Practices, Fragmentation | [34, 25] |
| European Union (EU) | The EU's proposed AI Act adopts a risk-based approach, tailoring regulations based on potential AI dangers. It categorises AI applications into unacceptable risk (outright prohibited), high risk (stringent requirements), low-to-minimal risk (less strict). Expected in 2024, the act aims to ensure AI safety and benefits. A notable section, Recital 107, requires developers of generative AI to transparently document the data used in model training, including copyrighted works. | Safety, Transparency of Data Use | [28, 30, 24] |
While some regulations exist for AI (e.g., EU's AI Act) it is crucial to note that these regulations do not necessarily translate into copyright regulations for GAI. Thus, a crucial aspect that emerges is the lack of specific copyright regulations or legislation tailored for GAI. This void leaves critical questions and research gap regarding copyright ownership of GAI-generated content and the implications of using copyrighted material and their quality (including trustworthiness) for GAI training. This also draws our attention to the challenge of determining the originality of the contents, and noting whether they are copyrighted or not? It is evident that the copyright of GAI is a complex phenomenon. Thus, the scope of the article is limited to uncover the copyright concerns for GAI, and determining the originality and copyright of the contents are subject to future research. It marks the necessity for governments to provide appropriate regulatory safeguards, delineate clear policies and guidelines on the permissible use and misuse of GAI along with repercussions for misuse. Thus, as we delve deeper into the next section, a more detailed examination of these copyright concerns surrounding GAI will uncover the complexities of authorship, ownership, fair use, and their impact on the creative landscape. Understanding these intricacies is vital to foster responsible data and GAI innovation while upholding individual rights and ethical considerations in the rapidly evolving world of GAI.
2.3 Legal Cases: Generative AI and Copyright Concerns
Navigating the intersection of GAI technology and copyright law is challenging. Copyright laws, as per [6], remain ambiguous and “technology neutral” in the face of emerging GAI technologies. GAI has substantially disrupted and transformed conventional content generation norms, thereby challenging established copyright precedents [38, 39]. Pertinent questions arise, such as whether employing potentially copyrighted material for GAI training constitutes copyright infringement, or whether the output generated by GAI possesses originality substantial enough for copyright protection. Additionally, the dilemma of identifying the authorship of content generated by GAI further complicates the legal landscape. We reviewed 10 copyright related legal cases to further unpack the copyright concerns. A concern (e.g., risk and value GAI) is a matter of interest to relevant stakeholders (e.g., developers, users, regulators). This work is informed by the risk management theory (RMT), which provides a model for risk informed decision making [72]. The RMT brings along fact-based and value-based aspects in the risk informed decision-making (e.g., evidence, knowledge base, broad risk evaluation, decision maker's review, decision). As per RMT theory, 10 legal cases—as evidence from public knowledge base—were inspected for a broader risk evaluation with a view to support decision maker's review and decision (e.g., GAI system developer's decision). RMT is relevant because the proposed CRC tool is designed to assist GAI developers in evaluating copyright concerns related to the inputs and outputs of their GAI systems, which will be further discussed in the subsequent sections of this article. Here, Table 4 presents the review of the selected legal cases and associated copyright concerns, serving as a critical reference point for identifying the need and foundation for the development of the proposed CRC tool. This process involved a thorough analysis of each case, considering factors such as the (1) clarity of copyright law in relation to GAI, (2) the precedent set by previous rulings, and (3) the evolving nature of technology and its use in content creation. It is important to note here that the rationale behind selecting these legal cases lies in their current relevance and their ability to encapsulate key copyright concerns for GAI systems. It is noteworthy that Australia, at this point, lacks publicly available GAI-related copyright legal cases.
Table 4.
| Case Number | Case Name | Copyright Concerns | Description | Relevant Concern Mapping |
|---|---|---|---|---|
| 1 | Kris Kashtanova's ‘Zarya of the Dawn’ Comic Book (2023) | Delineating AI-generated components | US Copyright Office acknowledged specific human-authored elements but refused protection for individual AI-generated images within the graphic novel [49]. | Transparency (clarifying AI vs. human contributions), Trust (in the copyright system's ability to adapt) |
| 2 | Telstra v Phone Directories Company (2010) | Human authorship and originality | Landmark Australian case asserting that a work must originate from a human author and possess a degree of creativity for copyright protection [51]. | Infringement (defining what can be copyrighted), Transparency (in the criteria for copyright) |
| 3 | Metallica v Napster, Inc. (2000) | Liability for secondary infringement of works | Napster provided the means for the peer-to-peer sharing of infringing musical works and sound recordings, raising concerns about secondary infringement liability [50]. | Infringement (liability for facilitating copyright violation) |
| 4 | IceTV Pty Ltd v Nine Network Australia Pty Ltd (2009) | Protects the specific expression of facts | Australian case where copyright protects the specific expression of facts, not the facts themselves, emphasising the importance of originality [53]. | Transparency (in what constitutes original expression) |
| 5 | Author's Guild v OpenAI (2023) | Authorisation or licensing requirement | Case dealt with the unauthorised, unlicensed use of works to train GAI models, highlighting the need for proper authorisation or licensing [54]. | Infringement (use without license), Trust (in respecting copyright laws) |
| 6 | Plaintiffs v Google (2023) | Compliance with copyright licensing in GAI system development | Challenges the legality of Google's AI systems training on copyrighted content without proper licensing, highlighting the need for compliance with intellectual property rights in the development of GAI technologies [57]. | Infringement (lack of licensing), Trust (in legal compliance) |
| 7 | Doe v GitHub et. Al (2022) | Adherence to licensing terms for open-source material | Use of open-source software by AI without adhering to licensing terms such as attribution, infringing copyright owners' rights [55]. | Infringement (ignoring licensing terms), Transparency (in open-source licensing) |
| 8 | Feist Publications v Rural Telephone Service Company (1991) | Originality and copyright protection of directories and fact-based works | US case that limits the scope of protection in fact-based works, questioning if the work AI is training on is based on facts or original expressions [52]. | Transparency (in the scope of copyright protection) |
| 9 | Getty Images v Stability AI (2023) | Unauthorised use of copyrighted material in GAI systems | Highlights the legal ramifications of unauthorised use of copyrighted material in GAI systems, emphasising proper licensing and adherence to intellectual property rights [56]. | Infringement (unauthorised use), Trust (in upholding copyright) |
| 10 | Jason Allen's ‘Théâtre D'opéra Spatial’ Artwork (2023) | Human authorship in AI-generated creative works | US Copyright Office ruled that the work lacks human authorship" and falls outside the purview of copyright law, which excludes works produced by non-humans [58]. | Transparency (in authorship criteria), Trust (in the system's recognition of human creators) |
In framing our analysis, it is essential to articulate the three central copyright concerns that inform this study: infringement, transparency, and trust. Copyright infringement is the unauthorised exploitation of copyrighted works, which violates the exclusive rights bestowed upon copyright holders [47]. An example of this is the commercial distribution of a copyrighted image without the consent of the copyright owner, constituting a breach of the owner's rights [38]. Transparency in copyright demands the forthright disclosure of copyright terms, licensing agreements, and authorship to ensure all involved parties understand their rights and duties [43]. A practical instance is when a software developer clearly states the licensing terms for an open-source project [4]. Trust in the copyright system is the confidence that creators, users, and the wider public place in the equitable application and governance of copyright rights [28]. This trust is exemplified by consumers engaging in lawful transactions that provide equitable remuneration to authors, thus upholding the system's fairness [16].
These concerns are intimately related to fundamental copyright concepts. Ownership is pivotal to the issue of infringement, as unauthorised use constitutes an infringement of the copyright owner's rights [36]. Fair use is a concept that intersects with both infringement and transparency, permitting the conditional use of copyrighted material without direct authorization, contingent upon clear and established guidelines [60]. Originality, a criterion for copyright protection, is vital in evaluating both infringement and transparency, as it helps to discern the uniqueness of a work [30]. The term “copyright of content” encompasses the breadth of works that are protected by copyright law and is at the heart of these three concerns [48]. Intellectual property rights, including copyrights, are the legal underpinnings that foster trust in the copyright system [61].
With these definitions in place, Table 4 presents a detailed examination of selected legal cases and the copyright issues they highlight. Each case is analysed with respect to its impact on infringement, transparency, and trust, offering an indispensable point of reference for the creation of the CRC tool and advancing the scholarly conversation on copyright in the era of GAI.
This section presented a review of the GAI systems, regulations, and standards, and 10 legal cases that draw our attention to the need for systemically identifying and assessing the GAI system copyright concerns related to their inputs and outputs. This provides motivation for the development of the proposed CRC tool, which is the focus of this article.
3 Research Method
The Design Science Research Methodology (DSRM) was specifically selected for its structured approach, which is highly effective in addressing practical problems through the creation and assessment of artefacts such as models, tools, and frameworks [23]. This methodology was favoured over traditional research methods, such as empirical studies or theoretical analysis (e.g., [70, 71]), as it allows for iterative development and evaluation of the proposed CRC Tool, ensuring the tool's relevance and utility for GAI developers and creators.
The CRC Tool is designed to assist GAI developers and creators in evaluating copyright concerns related to the inputs and outputs of their GAI systems. Our application of DSRM, as shown in Figure 2, involved a systematic process that included the identification of the problem, defining the objectives of the CRC Tool, its development, and a rigorous evaluation phase.
Fig. 2.
To evaluate the CRC Tool, we established a set of evaluation criteria that focus on its applicability—in particular its’ functionality and usability—in addressing copyright issues within GAI systems. These criteria were derived from the objectives defined in the initial stages of the DSRM process and informed by the analysis of GAI systems, pertinent regulations and standards, and insights from 10 legal case studies.
An “experimental scenario” was crafted to serve as a tangible case for applying the CRC Tool. This scenario provided a realistic context that allowed us to demonstrate how the tool functions in a real-world situation, contributing significantly to the evaluation process. During the evaluation, the scenario was used to test the tool's capabilities and assess its performance against the predefined criteria. This practical application highlighted the tool's strengths and potential areas for improvement.
The steps and procedures undertaken in the evaluation, through the lens of the experimental scenario, included the deployment of the CRC Tool in a web-based questionnaire format, the application of the tool within the scenario, and the subsequent analysis of the results. Each of these activities was designed to thoroughly test the tool and provide empirical evidence for its applicability. The detailed process of the CRC Tool's development and evaluation, as well as the implications of the findings, are further elaborated in the subsequent sections of this article.
4 The CRC Tool
The CRC Tool1 is a web-based tool for identifying and assessing the GAI system copyright concerns. The CRC Tool was conceptualised as an interactive educational instrument, aiming to inform GAI system developers about the nuances of copyright concerns in the realm of AI-generated content. Its inception lies in the recognition of the pressing need to bridge the gap between the rapidly advancing field of GAI and the specific legal complexities surrounding copyright, as identified by [4, 30, 35, 37, 57]. The development of the CRC Tool involved the research and identification of 10 key questions extracted from the copyright concerns highlighted in the 10 legal cases, regulations, legislation, and academic literature discussed in this article. These resources represent a vast majority of the copyright concerns up to November 2023 and encapsulate a spectrum of concerns faced in the realm of AI-generated content.
The CRC Tool is organised into five key components (Table 5, Figure 3): Users, Admin, Interface, Tally.so platform, and Research Analysis. The structure of these components was inspired by existing assessment tools, such as the privacy impact assessment tools used by governments [69], to ensure a comprehensive coverage of the assessment process. The choice of five components is based on the need to address the core functions of user interaction, system management, user experience, operational support, and ongoing research integration.
Table 5.
| Number | Component | Description |
|---|---|---|
| 1 | User | The User component signifies the individuals engaging with the system, offering input and interacting with its features. |
| 2 | Admin | The Admin component signfies the administrators who will manage and oversee the system's functionality and user interactions. |
| 3 | Interface | The Web Browser component symbolises the interface through which users access the system, enabling seamless interaction and input submission. |
| 4 | Tally.so | The online form tool that is easy to use and allows for customisation of relevant questions. |
| 5 | Research and Analysis | Informs the CRC Tool's adaptation by processing legal cases, regulations, legislation, and academic literature, ensuring accuracy and relevancy in the generated insights. |
Fig. 3.
First and foremost, the Users component (Component 1) engages GAI developers and creators, serving as the primary interface for their interaction with the tool. Following this, the Admin component (Component 2) oversees the tool's functionality and security, ensuring its smooth operation including its enhancement. The Interface component (Component 3) provides a user-friendly interaction and experience for GAI developers. Tally.so Platform (Component 4) provides the operational foundation for the CRC Tool, facilitating seamless interactions between users and the assessment process. The Research Analysis component (Component 5), the foundation of the tool, operates within an adaptation loop. It continuously integrates insights from ongoing research, informing updates to the tool's questions and ensuring its responsiveness to the evolving landscape of GAI advancements and copyright concerns. Together, these components create a comprehensive and dynamic CRC Tool.
It is important to note that the relationship between components four and five, shown in Figure 3, isn't arbitrary; it's a deliberate effort to encapsulate the multifaceted and emerging copyright concerns. Each question serves as a signal, revealing specific facets of copyright law pertinent to GAI systems. The core of the CRC Tool is the set of copyright questions, which have been identified based on the analysis of 10 legal cases. The selection of these cases is informed by their relevance to GAI and copyright concerns [4, 30]. Each legal case was analysed in detail to identify concerns such as authorship, ownership, copyright infringement, and ethical considerations. These identified concerns were subsequently translated into a set of questions incorporated into the CRC Tool.
The creation of these questions was informed by a synthesis of the legal cases with current academic literature and regulatory guidelines to ensure legal accuracy and practical relevance. Preliminary feedback from experts in copyright law, AI ethics, technology development and senior members of the research team were sought to validate the relevance and phrasing of the questions, aligning them with the broader community's understanding of the pressing copyright issues in the field of GAI.
Each question within the tool is included to mirror the legal intricacies presented in the legal cases, providing users with a structured framework to identify and assess copyright-related concerns within their GAI systems. Table 6 illustrates the connection between these CRC questions and the legal cases, reinforcing the rationale behind the inclusion of the 10 questions in the CRC Tool.
Table 6.
| Case Number | Case Name | Identified Copyright Concerns | Mapping to Related CRC Tool Questions | Relevant Concern Mapping |
|---|---|---|---|---|
| 1 | Kris Kashtanova's ‘Zarya of the Dawn’ Comic Book (2023) | US Copyright Office acknowledged specific human-authored elements but refused protection for individual AI-generated images within the graphic novel [49]. | To what extent is the AI output generated from sufficiently original and creative content? | Transparency (clarifying AI vs. human contributions), Trust (in the copyright system's ability to adapt) |
| 2 | Telstra v Phone Directories Company (2010) | Landmark Australian case asserting that a work must originate from a human author and possess a degree of creativity for copyright protection [51]. | How well have you determined the degree of human input and contribution to the AI generation process? | Infringement (defining what can be copyrighted), Transparency (in the criteria for copyright) |
| 3 | Metallica v Napster, Inc. (2000) | Napster provided the means for the peer-to-peer sharing of infringing musical works and sound recordings, raising concerns about secondary infringement liability [50]. | How effectively have you identified any existing copyrighted works that may have influenced the AI output? | Infringement (liability for facilitating copyright violation) |
| 4 | IceTV Pty Ltd v Nine Network Australia Pty Ltd (2009) | Australian case where copyright protects the specific expression of facts, not the facts themselves, emphasising the importance of originality [53]. | How well have you determined the degree of human input and contribution to the AI generation process? | Transparency (in what constitutes original expression) |
| 5 | Author's Guild v OpenAI (2023) | Case dealt with the unauthorised, unlicensed use of works to train GAI models, highlighting the need for proper authorisation or licensing [54]. | To what extent is the AI output generated from sufficiently original and creative content? | Infringement (use without license), Trust (in respecting copyright laws) |
| 6 | Plaintiffs v Google (2023) | Challenges the legality of Google's AI systems training on copyrighted content without proper licensing, highlighting the need for compliance with intellectual property rights in the development of GAI technologies [57]. | To what extent have you considered the potential moral and ethical implications of your AI-generated content? | Infringement (lack of licensing), Trust (in legal compliance) |
| 7 | Doe v GitHub et. Al (2022) | Use of open-source software by AI without adhering to licensing terms such as attribution, infringing copyright owners' rights [55]. | How effectively have you identified any existing copyrighted works that may have influenced the AI output? | Infringement (ignoring licensing terms), Transparency (in open-source licensing) |
| 8 | Feist Publications v Rural Telephone Service Company, Inc. (1991) | US case that limits the scope of protection in fact-based works, questioning if the work AI is training on is based on facts or original expressions [52]. | How well have you determined the degree of human input and contribution to the AI generation process? | Transparency (in the scope of copyright protection) |
| 9 | Getty Images v Stability AI (2023) | Highlights the legal ramifications of unauthorised use of copyrighted material in GAI systems, emphasising proper licensing and adherence to intellectual property rights [56]. | How effectively have you identified any existing copyrighted works that may have influenced the AI output? | Infringement (unauthorised use), Trust (in upholding copyright) |
| 10 | Jason Allen's ‘Théâtre D'opéra Spatial’ Artwork (2023) | US Copyright Office ruled that the work “lacks human authorship” and falls outside the purview of copyright law, which excludes works produced by non-humans [58]. | How well have you determined the degree of human input and contribution to the AI generation process? | Transparency (in authorship criteria), Trust (in the system's recognition of human creators) |
Designed to prompt in-depth analysis, each question begins with “How” or “To what extent,” encouraging respondents to thoroughly consider and address the copyright issues affecting their GAI systems. The phrasing of these questions is intentional, aiming to provoke comprehensive and reflective answers that engage users with the fundamental matters.
Users can respond to each copyright concerned question in the CRC Tool to identify the risk according to the predefined categories of low, moderate, or high risk as defined in Table 7. After responding to all 10 questions, the CRC Tool calculates an overall risk level indicator based on the responses. This indicated risk level, for the GAI system, is described along with possible mitigations or recommendations. While the RMT [72] provides the overall risk informed decision-making approach, the logic and details underpinning the risk levels are adapted from the NSW Government risk framework [15]. Thus, the proposed CRC Tool incorporates both the theory [72] and practice [15].
Table 7.
| Risk Level | Definition |
|---|---|
| High Risk | Systems demand immediate and decisive action to mitigate potential legal consequences [15]. |
| Moderate Risk | Systems require strategic enhancements to their processes and legal understanding, balancing urgency with strategic planning [15]. |
| Low Risk | Systems benefit from ongoing vigilance and proactive measures to avoid future complications [15]. |
The risk levels (high, moderate, and low) are defined (also in Table 7) as such: high-risk systems demand immediate and decisive action to mitigate potential legal consequences; moderate-risk systems require strategic enhancements to their processes and legal understanding, balancing urgency with strategic planning; low-risk systems benefit from ongoing vigilance and proactive measures to avoid future complications. Here, we used the NSW Government framework as an example; however, the proposed CRC Tool can be adopted and configured to include other frameworks as well.
Building upon the risk level definitions outlined in Table 7, the CRC Tool facilitates in the identification and assessment of potential legal and ethical implications associated with GAI systems. This assessment is crucial for developers and creators as they navigate the complex intersection between value and risk compliance with copyright laws (e.g., RMT [72]). The CRC Tool's structured approach allows users to systematically evaluate their GAI systems against each risk category, ensuring that all copyright concerns are thoroughly examined. The subsequent mapping of these risk levels to specific questions, as depicted in Table 8, provides a clear and actionable framework for identifying areas of concern. By categorising the risks as low, moderate, or high, the CRC Tool empowers users to prioritise their responses and focus on the most pressing issues that require attention. This prioritisation is essential for an effective assessment of copyright risks, and for fostering responsible practices in the development and use of GAI systems.
Table 8.
| CRC Question | Low Risk | Moderate Risk | High Risk |
|---|---|---|---|
| To what extent is the AI output generated from sufficiently original and creative content? | Demonstrates high originality and creativity, minimising the risk of copyright concerns. | Shows moderate originality, with potential for minor similarities to existing content. | Displays limited originality, raising potential copyright risks. |
| How well have you determined the degree of human input and contribution to the AI generation process? | Clear attribution and understanding of human-AI roles, reducing ambiguity. | Some attribution clarity, but roles may need further clarification. | Ambiguous attribution, leading to uncertainty about human-AI contributions. |
| How effectively have you identified any existing copyrighted works that may have influenced the AI output? | Thorough identification of potential influences, minimising copyright risks. | Some influences acknowledged, with potential for overlooked content. | Limited identification of potential influences, raising significant risks. |
| How comprehensively have you assessed the risks associated with using copyrighted data for training the AI model? | Comprehensive risk assessment and mitigation strategies in place. | Partial risk assessment with room for further analysis. | Limited risk assessment, posing potential legal and ethical concerns. |
| How well are you aware of the potential legal challenges related to AI-generated content? | In-depth awareness of legal challenges, ensuring proactive compliance. | Moderate awareness of legal challenges, with room for improvement. | Limited awareness of legal challenges, raising potential issues. |
| To what extent have you considered the potential moral and ethical implications of your AI-generated content? | Thorough consideration of ethical implications, guiding responsible AI development. | Some consideration of ethical implications, with potential for further analysis. | Limited consideration of ethical implications, posing ethical concerns. |
| How thoroughly have you reviewed relevant copyright laws and regulations in your jurisdiction? | In-depth review of copyright laws, ensuring compliance and minimising risks. | Moderate review of copyright laws, with potential for further understanding. | Limited review of copyright laws, raising potential legal issues. |
| How much are you using AI-generated content in a way that respects the rights of the original creators? | Strong adherence to rights of original creators, respecting intellectual property. | Some adherences, with potential for better alignment with creator rights. | Limited respect for creator rights, posing potential copyright violations. |
| How deeply have you considered the potential impact of AI-generated content on existing markets or industries? | Thorough consideration of market impact, with strategies to mitigate disruptions. | Some consideration, with room for further analysis of potential impacts. | Limited consideration of market impact, raising potential challenges. |
In line with RMT, the CRC Tool's methodology aligns with the proactive identification and assessment of potential risks. By integrating RMT into the CRC Tool, users are provided with a theoretical foundation that supports the systematic approach to risk assessment [72].
The CRC Tool's initial recommendations or proposed mitigation are intricately linked to the assessed risk levels of the GAI systems, ensuring a semi-tailored approach to each system's unique challenges. The logic behind these differentiated recommendations, shown in Table 8, lies in the varying degrees of risk and legal exposure. In essence, the recommendations are not one-size-fits-all; they can be tailored to address the specific challenges or risk levels, thereby ensuring a nuanced and effective response to the complex landscape of GAI and copyright law [15].
The recommendations embedded in the CRC Tool are derived from research and analysis of legal precedents, and AI regulations in the field of GAI and copyright law, as shown in Table 9. These initial recommendations are not only aligned with existing laws and AI regulations but also forward-looking, anticipating the future trajectory of GAI technology and its intersection with intellectual property rights [24, 27, 31, 34]. Further, it is important to note here that the CRC Tool is intended to be used as in initial self-assessment of a GAI system to identify and assess copyright concerns. This assessment, along with the generic recommendations provided by the CRC Tool to mitigate identified risks or copyright concerns, can be reviewed by a formal legal or GAI system review committee for more accuracy and greater details.
Table 9.
| Recommendations | Relevant Legal Cases | Relevant AI Regulations |
|---|---|---|
| Thoroughly Review GAI Outputs | All Cases | Australia: Emphasises transparency and fairness (AI assurance framework) [31]. EU: AI Act's transparency obligations for developers of generative AI models, as outlined in Recital 107 [24]. UK: Data Protection and Digital Information Bill, focusing on data privacy [26]. US: Importance of staying updated due to the fragmented regulatory landscape [34]. |
| Seek Professional Legal Advice | Cases where legal complexities and liabilities are highlighted (e.g., Case 3: Metallica v Napster, Inc., Case 5: Author's Guild v OpenAI, Case 7: Doe v GitHub et. Al, Case 9: Getty Images v Stability AI) | EU: AI Act's transparency obligations emphasise the importance of legal consultation [24]. UK: Importance of legal consultation for AI-generated content [26]. US: Fragmented regulations require legal expertise for compliance [34]. |
| Maintain Clear Documentation | Cases involving the need for clear authorship and licensing (e.g., Case 1: Kris Kashtanova's ‘Zarya of the Dawn’, Case 5: Author's Guild v OpenAI, Case 7: Doe v GitHub et. Al) | UK: Clarity in documentation aligns with UK's emphasis on data privacy [26]. Australia: Clear records of AI development processes align with transparency principles [27]. |
| Stay Updated on Copyright Laws | All Cases | Australia: Staying updated is essential to ensure compliance with evolving laws [27]. EU: AI Act's transparency obligations require ongoing awareness of legal requirements [24]. US: Fragmented regulations necessitate continuous awareness of copyright laws [34]. |
Table 10.
| Criterion Type | Description |
|---|---|
| Fit for Purpose | The CRC Tool accurately identifies copyright concerns in a variety of GAI systems, ensuring applicability and relevance within the ever-evolving GAI and copyright landscape. |
| Novelty | The CRC Tool serves as an innovative solution, bridging gaps in existing literature and resources, addressing ambiguities and challenges related to copyright concerns in GAI systems. |
| Ease of Use | The CRC Tool is designed to be easily accessible via web and user-friendly, making it a practical solution for addressing the complexities of GAI-related copyright issues. |
In summary, the CRC Tool provides an important development into the field of GAI systems and their copyright concerns. It enables users to conduct an initial risk assessment of their GAI system. It serves as an educational compass, helping GAI creators and developers to identify and assess copyright concerns, enabling them to make informed decisions and fostering a culture of responsible GAI development. Through this alignment, the CRC Tool is a starting point to further investigations of the complex intersection of GAI technology and copyright law.
5 CRC Tool Evaluation
The evaluation of the CRC Tool is performed through an “experimental” scenario, which is considered appropriate as per the DSRM research evaluation guidelines [24] and criterion types mentioned in Table 10 (Column 1). The criterion details, relevant to this research, were derived from the literature on GAI technology and copyright law [59, 60]. The literature analysis indicated common issues faced by GAI creators/developers and users when dealing with copyright concerns [61, 62], and the identification of key attributes that a tool must possess to effectively address these issues [63, 64].
The chosen criterion types – Fit for Purpose, Novelty, and Ease of Use – were selected from the DSRM for their collective ability to provide a reasonable assessment of the tool. The rationale behind selecting only these three criteria is based on their broad coverage of the essential aspects of the tool's functionality and its potential impact on the intended user (e.g., GAI developers). “Fit for Purpose” ensures that the tool meets the specific needs of identifying copyright concerns related to GAI systems, which is the primary objective of the CRC Tool. “Novelty” assesses the tool's contribution to the field, considering the lack of existing tools with similar capabilities, and its ability to offer new insights for identifying copyright concerns [65, 66]. “Ease of Use” is crucial for adoption and effective use, as a tool that is difficult to navigate or understand could hinder its practical application, regardless of its capabilities [65, 66].
Given the novelty of the CRC Tool in a landscape where similar tools are non-existent, this initial evaluation aims to establish a foundational benchmark or starting point for this important area of research. The CRC Tool is designed not only to fill the current gap but also to set a precedent for evaluating copyright concerns in GAI systems. The specific evaluation criteria employed are justified as follows:
By focusing on these three criteria, the evaluation encompasses the tool's applicability to real-world problems, its contribution to the field, and its usability, which are considered fundamental for the successful implementation and adoption of the CRC Tool [67, 68].
5.1 The Creation of the Experimental Scenario
To facilitate the evaluation of the CRC Tool, an experimental scenario was carefully designed using a scenario framework, as suggested in the literature [5]. The purpose of this scenario is to simulate a realistic and tangible context in which the CRC Tool can be applied and assessed. The scenario framework encompasses five distinct components, each serving a crucial role in the evaluation process. These components are visualised in Figure 4 and elaborated upon, which mirrors the structure outlined in Table 11.
Fig. 4.
Table 11.
| Number | Component | Description |
|---|---|---|
| 1 | Source | The source serves as the trigger for each scenario, representing the situation or event that prompts the need for copyright assessment and the use of the CRC Tool. The scenario was inspired by real legal cases to ensure their authenticity and applicability. By drawing from these cases, the source is grounded in actual copyright challenges faced by organisations and individuals in the GAI landscape. |
| 2 | Stimulus | The stimulus represents the specific issue or incident that arises from the source. In the context of these scenarios, the stimulus is the discovery of unlicensed copyrighted materials within AI-generated content. This stimulus is aligned with the copyright concerns highlighted in the legal cases, ensuring that the scenarios capture the essence of copyright infringements and compliance challenges. |
| 3 | Artifact | The artifact in the scenario is the CRC Tool. This Tool is employed to assess and evaluate copyright-related risks and to suggest next steps or recommendations for addressing them. By incorporating the CRC Tool as the artifact, the scenario provides a practical and structured means to indicate copyright concerns, mirroring real-world practices of using assessment tools to navigate copyright issues. The CRC Tool does not enforce compliance to specific standards or regulations, nor does it check against regulation databases; it serves to guide the developer in identifying areas that may require further legal scrutiny or action. |
| 4 | Response | The response component outlines the actions or measures taken in response to the stimulus. In these scenarios, the response includes the completion of the CRC Tool, and any subsequent steps taken based on its assessment/advice. This element reflects the proactive approach of organisations and individuals in addressing copyright challenges, as observed in the legal cases. |
| 5 | Response Measure | The response measure evaluates the effectiveness of the CRC Tool in identifying relevant or potential risks. It does not assess how well a company or individual addresses copyright concerns post-assessment. The focus is on the tool's capability for risk identification and assessment, in line with the original research question. The CRC Tool's objective is to assist developers in recognising potential copyright issues; it does not verify the originality of content, which remains the developer's responsibility. |
The concept of the scenario begins with a “Source” (Component 1), acting as the catalyst for the scenario's narrative. This element is derived from real-world legal cases, thereby ensuring that the scenario retains authenticity and relevance to actual copyright challenges encountered in the GAI domain. Proceeding from the source is the “Stimulus” (Component 2), a specific issue or incident that arises and necessitates copyright assessment. The choice of this stimulus is deliberate, reflecting the real-world copyright concerns that the legal cases exemplify, and thus it ensures that the scenario accurately encapsulates the nature of copyright infringement and compliance challenges.
At the heart of the scenario is the “Artifact” (Component 3), which, in this case, is the CRC Tool itself. The deployment of the CRC Tool in this scenario provides a structured approach for identifying and assessing copyright-related risks.The “Response” (Component 4) element details the course of action taken in light of the stimulus. Within the scenario, this involves the utilisation of the CRC Tool by the developer, resulting in the tool providing an indicative risk assessment output. Lastly, the “Response Measure” (Component 5) evaluates the outcomes of the actions taken. It determines the extent to which the CRC Tool managed to identify relevant or potential risks. This measure is a critical endpoint, as it aligns with the overarching goals of the CRC Tool: to identify and assess the copyright concerns and risks.
In summary, the experimental scenario, underpinned by the scenario framework and visualised in Figure 4, provides a structured and realistic environment for the rigorous assessment of the CRC Tool. The narrative formed by the five components serves as a comprehensive means to evaluate the tool's practicality and applicability in a manner that closely reflects real-world application and challenges.
5.2 The Experimental Scenario: CopiCode AI Training Controversy
To evaluate the CRC Tool2, we created a scenario featuring a hypothetical start-up named CopiCode, which is developing a GAI system. This scenario reflects the legal challenges start-ups may face when using datasets without proper authorisations, highlighting the balance between innovation and legal compliance in the GAI industry.
Real-life legal cases inform the scenario, providing a foundation of copyright compliance issues that enhance the scenario's realism. These cases offer insights into the legal challenges that GAI developers may encounter.
The CRC Tool's evaluation was conducted by applying it to the CopiCode scenario. This approach allowed for a practical examination of the tool's effectiveness in identifying potential copyright risks within a realistic context. The tool's performance was assessed to determine its usefulness for GAI developers in a scenario that resembles the challenges they could face.
5.3 Application of the CRC Tool for the CopiCode AI Training Controversy Scenario
Following the establishment of the CopiCode AI Training Controversy scenario, the CRC Tool's evaluative capabilities were put to the test. The scenario served as a testing ground for the CRC Tool's functionality. The hypothetical nature of CopiCode, exemplifies the tension between value and risk legal compliance – a balance that is critical in the rapidly evolving GAI sector.
The tool's comprehensive assessment covered various dimensions of copyright concerns and risk management, with insights informed by real-world legal cases to ensure practical relevance for GAI developers. As detailed in Table 12, the CRC Tool analysed CopiCode's copyright strategies through targeted questions.
Table 12.
| CRC Tool Question | CopiCode Response | Explanation |
|---|---|---|
| To what extent is the AI output generated from sufficiently original and creative content? | Moderate (B) | CopiCode's AI-generated code exhibits moderate originality but may have minor similarities to existing content. This suggests there is room for improvement in terms of originality. |
| How well have you determined the degree of human input and contribution to the GAI generation process? | Moderate (B) | While there is some attribution clarity, roles may need further clarification, indicating a moderate level of understanding regarding human-AI contributions. |
| How effectively have you identified any existing copyrighted works that may have influenced the GAI output? | Low (C) | CopiCode has limited identification of potential influences, raising significant copyright risks. This suggests the need for better identification and management of copyrighted content. |
| How comprehensively have you assessed the risks associated with using copyrighted data for training the GAI model? | High (A) | CopiCode has conducted a comprehensive risk assessment and has mitigation strategies in place, demonstrating a high level of preparedness regarding copyright risks. |
| How well are you aware of the potential legal challenges related to AI-generated content? | High (A) | CopiCode shows in-depth awareness of legal challenges, ensuring proactive compliance with legal requirements. |
| To what extent have you considered the potential moral and ethical implications of your AI-generated content? | Moderate (B) | There is some consideration of ethical implications, with potential for further analysis. While ethical concerns are acknowledged, they may benefit from more thorough examination. |
| How thoroughly have you reviewed relevant copyright laws and regulations in your jurisdiction? | High (A) | CopiCode has conducted an in-depth review of copyright laws, ensuring compliance and minimising risks. |
| How much are you using AI-generated content in a way that respects the rights of the original creators? | Moderate (B) | CopiCode demonstrates some adherence to the rights of original creators, with potential for better alignment with creator rights. |
| How deeply have you considered the potential impact of AI-generated content on existing related markets or industries? | Moderate (B) | There is some consideration of market impact, with room for further analysis of potential impacts. CopiCode has acknowledged the importance of market impact but may need a more detailed assessment. |
| How extensively have you sought legal advice or consulted professionals for copyright-related concerns related to AI-generated content? | Moderate (B) | CopiCode has engaged in moderate consultation with legal professionals, indicating a willingness to seek legal insights. |
The findings from the CRC Tool's assessment, indicated that CopiCode's AI-generated code had moderate originality but also bore similarities to existing works, suggesting a need for more innovation. The company's grasp of human and AI contributions to content creation was also moderate, calling for clearer attribution methods.
CopiCode's ability to identify copyrighted influences in their GAI output was rated low, highlighting a significant area for improvement. Conversely, their risk assessment practices, and legal awareness were highly rated, showing a proactive stance on copyright risks. Ethical considerations, respect for original creators' rights, and the potential market impact of their AI content were all rated moderate, suggesting an understanding of these factors but also the need for deeper analysis and strategic planning. Lastly, their engagement with legal professionals was moderate, indicating some reliance on expert advice but also the possibility for more comprehensive legal consultation.
The CRC Tool's assignment of a “Moderate Risk” score to the CopiCode scenario, as depicted in Figure 5, is a testament to the tool's analytical framework. This score recognises CopiCode's proactive measures in addressing legal compliance, while also suggesting areas where their strategies could be improved. The “Moderate Risk” classification is not an indication of failure but rather a call to action, signalling areas where CopiCode can enhance their practices to better safeguard their GAI system against potential legal infringements.
Fig. 5.
In response to the identified risks in Figure 5, Figure 6 presents a set of generic recommendations aimed at guiding CopiCode through the challenges it faces. While these recommendations provide a starting point for improvement, they are intended to be followed up with a consultation with a legal advisor to ensure integration into the company's operational framework. This approach ensures that CopiCode can develop more robust systems and protocols to detect and prevent the use of copyrighted material, ultimately reducing the risk of infringement.
Fig. 6.
The recommendations outlined in Figure 6 are indicative of the CRC Tool's alignment with the purpose of aiding GAI developers in identifying, assess and mitigating copyright risks effectively. By adhering to these strategies, GAI developers like CopiCode can enhance their legal compliance, foster innovation responsibly, and maintain the integrity of their creative processes.
The analysis underscores the CRC Tool's practical relevance, offering general recommendations that are both innovative and applicable to various GAI systems. By linking the hypothetical scenario to real-world legal cases, the evaluation provides actionable insights for organisations and individuals to address copyright concerns effectively. Moreover, the Tool's user-friendly interface and the website's seamless internet accessibility are key factors in its ease of use, streamlining the identification and assessment of copyright concerns for developers and creators, and ensuring the seamless integration of ethical and legal considerations into the GAI development process prove it to be an invaluable asset for the GAI industry.
6 Discussion and Conclusion
This article has presented significant work concerning GAI systems and the associated copyright concerns. A key insight from this research is the lack of GAI-specific regulations and the reliance on “technology neutral” copyright laws, which may lead to an increase in copyright infringements and violations by GAI systems [6]. This issue poses a threat not only to the rights of original content creators but also to the integrity of the legal framework governing intellectual property rights. Consequently, this research was motivated by the critical question: how can we identify and assess the copyright concerns of GAI systems? In response, this article proposed a research-based CRC Tool designed to enable developers to self-assess the copyright health of their GAI systems.
The development of the CRC Tool represents a significant stride toward addressing the copyright infringements and violations evidenced in publicly available legal cases across Australia, the US, UK, and EU. The contributions and insights of this research are multifaceted:
Academically, the CRC Tool elucidates the ambiguities and complexities at the intersection of GAI and copyright laws, laying the groundwork for a practical tool that will inform future research. The analysis of current AI regulations and standards, along with the exploration of GAI's copyright concerns, enriches the academic discourse.
Practically, the CRC Tool provides actionable insights and recommendations for GAI developers and creators. It offers a structured framework to navigate the intricate landscape of copyright, empowering users to make informed decisions. As a proactive measure, the CRC Tool aids in identifying and assessing copyright concerns, thereby promoting a culture of responsible innovation.
Socially, this research enhances ethical awareness within the GAI community. By advocating for education and the application of the CRC Tool, it encourages a commitment to legal compliance and ethical development, ensuring that GAI technologies are not only innovative but also ethically responsible.
Technologically, the CRC Tool serves as a vital guide for GAI system creators and developers, aligning technological innovations with existing legal frameworks. This alignment is crucial not only to avoid legal complications but also to support the development of responsible and legally compliant GAI systems.
While this research addresses a significant gap, it is also important to recognise its limitations and the scope of the proposed solution. The CRC Tool is based on specific instances of copyright infringements or violations documented in public legal cases from selected regions. Its applicability may be limited by the rapidly evolving nature of GAI technology and the potential for new copyright challenges not yet seen in existing legal cases. As the field of GAI continues to grow, it is expected that the CRC Tool and its foundational questions will adapt to incorporate new insights from emerging cases.
Moreover, it is crucial to consider that different fields and regions may have unique regulations, standards, and legal precedents related to GAI and copyright. The configuration and applicability of the CRC Tool may need to be tailored to the specific legal contexts of various countries, industries, and creative domains, highlighting the importance of local and sector-specific considerations alongside the Tool's guidance.
The initial evaluation of the research findings is based on the functionality of the CRC Tool, assessed against three key criteria: novelty, fit for purpose, and ease of use. The newly proposed research-based CRC Tool appears to address the research gap identified in Section 2. The user-friendly questionnaire-based web interface is suitable for the task of identifying and assessing GAI system copyright concerns, as demonstrated by the evaluation scenario results. Future work should focus on expanding the CRC Tool to cover a wider array of legal cases and integrating automated updates from academic literature and legal cases, ensuring that the CRC Tool remains relevant and effective in the field of GAI and copyright law.
Footnotes
1
The CRC Tool is available here: https://tally.so/r/wdalqD
2
The Copyright Checker (CRC) Tool is available here: https://tally.so/r/wdalqD
References
[1]
Accenture and Elevandi. 2023. Responsible AI in Finance: Navigating the Ethics of Generative AI. Retrieved 02-Nov-2023 from https://www.aiunplugged.io/wp-content/uploads/2023/10/Responsible-AI-In-Finance-Navigating-the-Ethics-Of-Generative-AI.pdf
[2]
ADL Center for Technology and Society. 2023. Six Pressing Questions We Must Ask About Generative AI | ADL. www.adl.org. Retrieved 02-Nov-2023 from https://www.adl.org/resources/blog/six-pressing-questions-we-must-ask-about-generative-ai
[3]
L. Alzubaidi, J. Zhang, A. J. Humaidi, A. Al-Dujaili, Y. Duan, O. Al-Shamma, J. Santamaría, M. A. Fadhel, M. Al-Amidie, and L. Farhan. 2021. Review of deep learning: concepts, CNN architectures, challenges, applications, future directions. Journal of Big Data 8, 1 (2021), 1--74. DOI:
[4]
Australian Government Attorney-General's Department. 2023. Issue summary paper: Artificial Intelligence (AI) and copyright. 2023. In Australian Government. Retrieved 02-Nov-2023 from https://www.ag.gov.au/sites/default/files/2023-09/issue-summary-paper-on-ai-and-copyright.PDF
[5]
L. Bass, P. Clements, and R. Kazman. 2021. Software Architecture in Practice (4th ed.). O'Reilly.
[6]
T. de Sousa and S. Lynch. 2023. WaiTalk 6: Personal Data and AI (Privacy by Design). Women in AI Summit 2023 Sydney.
[7]
P. Daugherty, B. Ghosh, K. Narain, L. Guan, and J. Wilson. 2023. Generative AI in Business. Www.accenture.com. Retrieved 02-Nov-2023 from https://www.accenture.com/es-es/insights/technology/generative-ai
[8]
Y. K. Dwivedi, N. Kshetri, L. Hughes, E. L. Slade, A. Jeyaraj, A. K. Kar, A. M. Baabdullah, A. Koohang, V. Raghavan, M. Ahuja et al. 2023. So what if ChatGPT wrote it? Multidisciplinary perspectives on opportunities, challenges and implications of generative conversational AI for research, practice and policy. International Journal of Information Management, 71 (0268-4012), 102642. DOI:
[9]
K. Firth-Butterfield. 2023. AI Leaders Call for Pause in Systems Training, Plus Other AI Stories to Read This Month. World Economic Forum. Retrieved from https://www.weforum.org/agenda/2023/03/ai-leaders-call-for-pause-plus-other-ai-stories-to-read-this-month/
[10]
A. Q. Gill. 2022. Adaptive Enterprise Architecture as Information: Architecting Intelligent Enterprises. World Scientific Publishing Company.
[11]
R. Golden. 2023. Most Employees Using ChatGPT, Other AI Tools for Work Aren't Telling Their Bosses. HR Dive. Retrieved 02-Nov-2023 from https://www.hrdive.com/news/employees-using-chatgpt-ai-for-work-not-telling-bosses/642166/
[12]
J. M. Helm, A. M. Swiergosz, H. S. Haeberle, J. M. Karnuta, J. L. Schaffer, V. E. Krebs, A. I. Spitzer, and P. N. Ramkumar. 2020. Machine learning and artificial intelligence: definitions, applications, and future directions. Current Reviews in Musculoskeletal Medicine 13, 1 (2020), 69–76. DOI:
[13]
J. L. McCarthy, M. L. Minsky, N. Rochester, and C. E. Shannon. 1955. A proposal for the Dartmouth summer research project on artificial intelligence. Retrieved August 31 from http://jmc.stanford.edu/articles/dartmouth/dartmouth.pdf
[14]
S. Noy and W. Zhang. 2023. Experimental evidence on the productivity effects of generative artificial intelligence. SSRN Electronic Journal. DOI:
[15]
NSW Government. 2022. Enterprise-Wide Risk Management. NSW Government. Retrieved 02-Nov-2023 from https://www1.health.nsw.gov.au/pds/ActivePDSDocuments/PD2022_023.pdf
[16]
S. Pawar. 2023. WaiTalk 8: Future of Technology Responsible AI in the Era of Generative AI. Women in AI Summit 2023 Sydney.
[17]
I. Rahwan, M. Cebrian, N. Obradovich, J. Bongard, J.-F. Bonnefon, C. Breazeal, J. W. Crandall, N. A. Christakis, I. D. Couzin, M. O. Jackson et al. 2019. Machine behaviour. Nature 568, 7753 (2019), 477–486. DOI:
[18]
P. N. Ramkumar, J. M. Karnuta, S. M. Navarro, H. S. Haeberle, R. Iorio, M. A. Mont, B. M. Patterson, and V. E. Krebs. 2019. Preoperative prediction of value metrics and a patient-specific payment model for primary total hip arthroplasty: development and validation of a deep learning model. The Journal of Arthroplasty 34, 10 (2019), 2228–2234.e1. DOI:
[19]
P. Stone, R. Brooks, E. Brynjolfsson, R. Calo, O. Etzioni, G. Hager, J. Hirschberg, S. Kalyanakrishnan, E. Kamar, S. Krauset al. 2016. Artificial intelligence and life in 2030. One Hundred Year Study on Artificial Intelligence: Report of the 2015- 2016 Study Panel, Stanford University, Stanford, CA. Retrieved from http://ai100.stanford.edu/2016-report
[20]
Z. Tekic and J. Füller. 2023. Managing innovation in the era of AI. Technology in Society 73 (2023), 102254. DOI:
[21]
L. Zhang, K. Li, and J. Liu. 2021. An information diffusion model based on explosion shock wave theory on online social networks. Applied Sciences 11, 21 (9996). DOI:
[22]
OECD. 2023. Updates to the OECD's definition of an AI system explained. Retrieved 02-Nov-2023 from https://oecd.ai/en/wonk/ai-system-definition-update
[23]
A. Hevner, S. March, J. Park, and S. Ram. 2004. Design science in information systems research. MIS Quarterly 28, 1 (2004), 75–105.
[24]
European Commission. 2021. The Act, The Artificial Intelligence Act.
[25]
D. Braue. 2022. Bill of Rights Defines Five Rules of Ethical AI. Information Age. Retrieved 02-Nov-2023 from https://ia.acs.org.au/article/2022/bill-of-rights-defines-five-rules-of-ethical-ai.html
[26]
M. Davies and M. Birtwistle. 2023. Regulating AI in the UK. Www.adalovelaceinstitute.org. Retrieved from https://www.adalovelaceinstitute.org/report/regulating-ai-in-the-uk/
[27]
Department of Industry, Science and Resources. 2022. Australia's AI Ethics Principles. Industry.gov.au. Retrieved from https://www.industry.gov.au/publications/australias-artificial-intelligence-ethics-framework/australias-ai-ethics-principles
[28]
S. Feingold. 2023. On Artificial Intelligence, Trust is a Must, Not a Nice to Have,” One Lawmaker Said. #AI. World Economic Forum. Retrieved from https://www.weforum.org/agenda/2023/06/european-union-ai-act-explained/
[29]
Gilbert + Tobin. 2022. Artificial Intelligence Work Group Project Australia | G+T. Www.gtlaw.com.au. Retrieved from https://www.gtlaw.com.au/knowledge/artificial-intelligence-work-group-project-australia
[30]
P. B. Hugenholtz and J. P. Quintais. 2021. Copyright and artificial creation: Does EU copyright law protect AI-Assisted output? IIC - International Review of Intellectual Property and Competition Law 52, 9 (2021), 1190--1216. DOI:
[31]
NSW Department of Customer Service. 2022. NSW Artificial Intelligence Assurance Framework | Digital.NSW. www.digital.nsw.gov.au. Retrieved 02-Nov-2023 from https://www.digital.nsw.gov.au/policy/artificial-intelligence/nsw-artificial-intelligence-assurance-framework
[32]
I. Rahwan, M. Cebrian, N. Obradovich, J. Bongard, J.-F. Bonnefon, C. Breazeal, J. W. Crandall, N. A. Christakis, I. D. Couzin, M. O. Jackson et al. 2019. Machine behaviour. Nature 568, 7753 (2019), 477–486. DOI:
[33]
H. Roberts, A. Babuta, J. Morley, C. Thomas, M. Taddeo, and L. Floridi. 2023. Artificial intelligence regulation in the United Kingdom: A path to good governance and global leadership? Internet Policy Review 12, 2 (2023), 1--31. Retrieved from https://policyreview.info/articles/analysis/artificial-intelligence-regulation-united-kingdom-path-good-governance
[34]
The White House. 2022. Blueprint for an AI Bill of Rights. The White House. Retrieved 02-Nov-2023 from https://www.whitehouse.gov/ostp/ai-bill-of-rights/
[35]
M. Sag. 2023. Senate Judiciary Committee Holds Hearing on AI and Copyright — 07/12/23. Www.youtube.com; CNBC Television. Retrieved from https://www.youtube.com/watch?v=uoCJun7gkbA
[36]
R. Matulionyte. 2021. Australian copyright law impedes the development of artificial intelligence: What are the options?’. IIC - International Review of Intellectual Property and Competition Law 52, 4 (2021), 417–443.
[37]
G. Appel, J. Neelbauer, and D. A. Schweidel. 2023. Generative AI Has an Intellectual Property Problem, Harvard Business Review, Harvard Business Publishing.
[38]
T. Golder, J. Gleeson, and B. Van Slyke. 2012. I, Object!: A landmark case in relation to software copyright. Journal for the Australian and New Zealand Societies for Computers and the Law 83 (2012), 1--5.
[39]
T. Pistorius and J. Jordaan. 2022. Whose data is it anyway? Copyright protection of databases and big data through the looking glass. Journal for the Australian and New Zealand Societies for Computers and the Law 94 (2022), 1--12.
[40]
R. Green. 2021. Navigating the AI transformation journey. Journal for the Australian and New Zealand Societies for Computers and the Law 93 (2021).
[41]
E. Wright, J. Zhou, D. Lindsay, L. Przhedetsky, F. Chen, and A. Davison. 2022. SMEs and explainable AI: Australian case studies. Journal for the Australian and New Zealand Societies for Computers and the Law 94 (2022), 1--22.
[42]
R. Matulionyte. 2022. A study on explainable AI in healthcare: A brief report. Journal for the Australian and New Zealand Societies for Computers and the Law 94 (2022), 1--5.
[43]
H. P. Olsen, T. T. Hildebrandt, C. Wiesener, M. S. Larsen, and A. W. Ammitzbøll Flügge. 2023. The right to transparency in public governance: Freedom of information and the use of artificial intelligence by public agencies. Digital Government: Research and Practice 5, 1 (2023), 1--15.
[44]
M. Safaei and J. Longo. 2021. The end of the policy analyst? Testing the capability of artificial intelligence to generate plausible, persuasive, and useful policy analysis. Digital Government: Research and Practice 5, 1 (2021), 1--35.
[45]
R. C. Presuel and J. M. Martinez Sierra. 2024. The adoption of artificial intelligence in bureaucratic decision-making: A weberian perspective. Digital Government: Research and Practice 5, 1 (2024), 1--20.
[46]
T. D. Huynh, N. Tsakalakis, A. Helal, S. Stalla-Bourdillon, and L. Moreau. 2021. Addressing regulatory requirements on explanations for automated decisions with provenance—a case study. Digital Government: Research and Practice 2, 2 (2021), 1–14.
[47]
Arts Law Centre of Australia. 2010. Copyright Information Sheet. Arts Law Centre of Australia. Retrieved 02-Nov-2023 from https://www.artslaw.com.au/information-sheet/copyright
[48]
Australian Libraries and Archives. 2023. A Short Introduction to Copyright | Australian Libraries and Archives Copyright Coalition. A Short Introduction to Copyright. Retrieved 02-Nov-2023 from https://alacc.org.au/a-short-introduction-to-copyright/
[49]
R. J. Kasunic. 2023. Zarya of the Dawn. United States Copyright Office. Retrieved 02-Nov-2023 from https://www.copyright.gov/docs/zarya-of-the-dawn.pdf
[50]
Metallica v. Napster, Inc., (United States District Court, N.D. California March 5, 2000). Retrieved 02-Nov-2023 from https://casetext.com/case/metallica-v-napster-inc/
[51]
J. Henderson and B. Coogan. 2010. No Copyright Protection for Telephone Directories - Telstra v Phone Directories Company - Copyright - Australia. Www.mondaq.com. Retrieved 02-Nov-2023 from https://www.mondaq.com/australia/copyright/94168/no-copyright-protection-for-telephone-directories-telstra-v-phone-directories-company
[52]
Feist Publications v Rural Telephone Service Company, Inc., 340 Justia Law \(\S\) 499 \(\_\!\_\!\_\) (U.S. Supreme Court 1991). Retrieved 02-Nov-2023 from https://supreme.justia.com/cases/federal/us/499/340/
[53]
M. Zanker and D. Webber. 2009. Channel Nine's Copyright Case Against IceTV Goes to Water. Davies Collison Cave. Retrieved 02-Nov-2023 from https://dcc.com/news-and-insights/channel-nines-copyright-case-against-icetv-goes-to-water/
[54]
Author's Guild. 2023. Author's Guild v OpenAI. Class Action Complaint. Author's Guild. Retrieved 02-Nov-2023 from https://authorsguild.org/app/uploads/2023/09/Authors-Guild-OpenAI-Class-Action-Complaint-Sep-2023.pdf
[55]
J. Vincent. 2022. The Lawsuit that Could Rewrite the Rules of AI Copyright. The Verge. Retrieved 02-Nov-2023 from https://www.theverge.com/2022/11/8/23446821/microsoft-openai-github-copilot-class-action-lawsuit-ai-copyright-violation-training-data
[56]
J. Vincent. 2023. Getty Images is Suing the Creators of AI art Tool Stable Diffusion for Scraping its Content. The Verge. Retrieved 02-Nov-2023 from https://www.theverge.com/2023/1/17/23558516/ai-art-copyright-stable-diffusion-getty-images-lawsuit
[57]
Clarkson Law Firm. 2023. Plantiffs v Google. Class Action Lawsuit. Thomson Reuters. Retrieved 02-Nov-2023 from https://fingfx.thomsonreuters.com/gfx/legaldocs/myvmodloqvr/GOOGLE%20AI%20LAWSUIT%20complaint.pdf
[58]
S. V. Wilson, M. Strong, and J. Rubel. 2023. Second Request for Reconsideration for Refusal to Register Théâtre D'opéra Spatial. United States Copyright Review Board. Retrieved 02-Nov-2023 from https://www.copyright.gov/rulings-filings/review-board/docs/Theatre-Dopera-Spatial.pdf
[59]
F. Pasquale. 2019. The Black Box Society: The Secret Algorithms that Control Money and Information. Harvard University Press.
[60]
P. Samuelson. 2019. Allocating ownership rights in computer-generated works. University of Pittsburgh Law Review 80, 4 (2019), 1–48.
[61]
P. S. Menell and S. Scotchmer. 2007. Intellectual property law. In Handbook of Law and Economics. A. M. Polinsky and S. Shavell, (Eds.), Vol. 2, Elsevier.
[62]
M. A. Lemley. 2018. IP in a world without scarcity. New York University Law Review 90, 3 (2018), 460–515.
[63]
L. Lessig. 2006. Code: Version 2.0. Basic Books.
[64]
G. Sartor and M. Viola de Azevedo Cunha. 2010. The Italian Google-case: Privacy, freedom of speech and responsibility of providers for user-generated contents. International Journal of Law and Information Technology 18, 4 (2010), 356–378.
[65]
D. A. Norman. 2013. The Design of Everyday Things: Revised and Expanded Edition. Basic Books.
[66]
N. Bostrom and E. Yudkowsky. 2014. The ethics of artificial intelligence. In The Cambridge Handbook of Artificial Intelligence. W. Ramsey and K. Frankish (Eds.), Cambridge University Press.
[67]
J. Nielsen. 1994. Usability Engineering. Academic Press.
[68]
B. Shneiderman and C. Plaisant. 2010. Designing the User Interface: Strategies for Effective Human-Computer Interaction (5th ed.). Addison-Wesley.
[69]
OAIC. 2023. Privacy Impact Assessment Tool. OAIC. Retrieved from https://www.oaic.gov.au/privacy/privacy-guidance-for-organisations-and-government-agencies/privacy-impact-assessments/privacy-impact-assessment-tool
[70]
A. Q. Gill. 2015. Adaptive enterprise architecture driven agile development. In Proceedings of the International Conference on Information Systems Development, ISD 2015.
[71]
A. Qumer and B. Henderson-Sellers. 2007. Construction of an agile software product-enhancement process by using an agile software solution framework (ASSF) and situational method engineering. In Proceedings of the 31st Annual International Computer Software and Applications Conference (COMPSAC 2007). IEEE, 1, 539–542.
[72]
S. O. Hansson and T. Aven. 2014. Is risk analysis scientific? Risk Analysis 34, 7 (2014), 1173–1183.
Index Terms
- Towards the Development of a Copyright Risk Checker Tool for Generative Artificial Intelligence Systems
Recommendations
Protection of Copyrights in the Era of Generative Artificial Intelligence
Progress in Artificial IntelligenceAbstractThe historical evolution of terms such as artificial intelligence and related concepts has perpetuated a misconception conflating machine intelligence with human cognition. Recent strides in generative AI technology have exacerbated this ...
Artificial intelligence and moral rights
AbstractWhether copyrights should exist in content generated by an artificial intelligence is a frequently discussed issue in the legal literature. Most of the discussion focuses on economic rights, whereas the relationship of artificial intelligence and ...
Ethics in the Use of Artificial Intelligence in the Media
HCI International 2024 – Late Breaking PapersAbstractThe use of artificial intelligence and robotics in the media is no longer a novelty and is rapidly entering all areas of life, including the media. Now, a robot can replace the news anchor, the moderator of TV shows, and more. Artificial ...
Comments
Please enable JavaScript to view thecomments powered by Disqus.Information & Contributors
Information
Published In
December 2024
114 pages
Copyright © 2024 Copyright held by the owner/author(s).
This work is licensed under a Creative Commons Attribution-NoDerivs International 4.0 License
Publisher
Association for Computing Machinery
New York, NY, United States
Publication History
Published: 16 December 2024
Online AM: 08 November 2024
Accepted: 20 October 2024
Revised: 14 September 2024
Received: 27 December 2023
Published in DGOV Volume 5, Issue 4
Check for updates
Author Tags
Qualifiers
- Research-article
Contributors
Other Metrics
Bibliometrics & Citations
Bibliometrics
Article Metrics
- 0Total Citations
- 547Total Downloads
- Downloads (Last 12 months)547
- Downloads (Last 6 weeks)179
Reflects downloads up to 03 Mar 2025
Other Metrics
Citations
View Options
Login options
Check if you have access through your login credentials or your institution to get full access on this article.
Sign in