INTRODUCTION
Generative Artificial Intelligence (GenAI) represents a transformative advancement in contemporary technology. GenAI enables the production of novel creative content, including text, images, audio, code, or video, surpassing the capabilities of earlier AI systems.1 While early forms of AI emerged in the mid-20th century, recent developments in generative models have significantly expanded its creative and functional capabilities. This technology embodies a type of artificial intelligence engineered to generate high-quality text, images, videos, audio, and synthetic data within seconds. A major breakthrough occurred in 2014 with the development of Generative Adversarial Networks (GANs), which enabled generative AI to advance to a level where it could produce images, videos, and audio that closely emulate real-world recordings. At present, generative AI plays a crucial role in ChatGPT and its variants.2
A. AI and Generative AI (GenAI)
Although AI and generative AI (GenAI) are both intelligent systems, they perform different functions. Traditional AI focuses on assessing pre-existing data to tackle problems or execute tasks, such as spam detection and facial recognition. Alternatively, GenAI is capable of producing completely novel data, such as generating images from textual descriptions or creating innovative text formats. Traditional AI typically operates on predefined datasets to analyse and predict outcomes. In contrast, GenAI generates novel outputs by learning patterns from large-scale datasets, employing data to generate unique outputs, including text, code, images, and video.3
Developing a GenAI model requires three fundamental components:
• Extensive datasets: Generative models need substantial quantities of data to determine patterns and yield new content efficiently.
• Advanced deep learning architectures: Approaches such as Generative Adversarial Networks (GANs) or Transformers are frequently employed to optimize these models.
• Computational resources: The training of generative models can be resource-demanding, calling for robust hardware structures such as Graphics Processing Units (GPUs).4
GenAI is widely utilized across multiple industries: in financial services for detecting fraud, underwriting loans, and optimizing portfolios; in retail and e-commerce for tailored marketing and customer support chatbots; in healthcare and pharmaceuticals for drug discovery, diagnosis, and personalized treatment; in IT and DevOps for generating code, fixing bugs, and automating tests; in manufacturing for designing products and forecasting supply chains; and in media and telecommunications for synthesising content, personalizing experiences, and optimizing networks, with new applications emerging in synthetic data, automated user interface and user experience, and three-dimensional simulations.5
The UN Trade and Development (UNCTAD) report shows that AI market is anticipated to grow tremendously, increasing from approximately 40 billion USD in 2022 to 1.3 trillion USD by 20326. According to Goldman Sachs, global investments in AI are expected to approach 200 billion USD by 2025, driven by the swift uptake of services such as ChatGPT, which achieved 100 million users within a mere two months.7
GenAI has progressed from basic forecasting and pattern recognition to the active generation of nuanced text, images, and even entire simulated environments. This advancement is accomplished by leveraging extensive and varied datasets, which facilitate the perception of creativity in its outputs.
B. GenAI for authentic creativity?
Mark Twain has famously stated “There is no such thing as a new idea. We simply take a lot of old ideas and put them into a sort of mental kaleidoscope.” 8 Creativity is not merely the manifestation of magical inspiration; rather, it involves the thoughtful integration of known ideas, patterns, and concepts in innovative approaches, thus, what we perceive as original or innovative is often simply a new composition of pre-existing elements.9
In his book Where Good Ideas Come From, Steven Johnson 10 presents the concept of the “adjacent possible,” positing that innovation typically arises from the novel combination of pre-existing ideas. This perception highlights the transformative impact of generative AI on creativity: by harnessing an extensive database, it is able to blend and reconfigure patterns to produce original outputs at an extent and pace that far exceed human abilities.11
Generative AI is carving a niche for itself in the heart of creative and artistic expression. From the strokes of a painter’s brush to the cadence of a musician’s chord and the richness of a writer’s prose, we are now being driven to reconsider our conception and creation of art.12
C. Intersection of law and GenAI
Generative AI systems secure patterns from comprehensive and diverse datasets, allowing them to generate human-like content expeditiously and at a low cost, enabling ordinary users to create in mere minutes what previously required hours of effort. As of mid-2024 McKinsey global survey, approximately 65% of organizations have integrated GenAI into at least one business function,13 and Ernst & Young project that productivity enhancements driven by GenAI could enhance global GDP by an estimated USD 1.2–2.4 trillion over the coming decade.14
Artificial Intelligence is now a key constituent of business operations and is set to catalyze major developments. However, its development and utilization have sparked intense debates in the field of intellectual property law. This is critically relevant when it comes to safeguarding AI-generated content, determining the ownership of rights related to these creations, and the risk of contravening existing intellectual property rights during the training or use of generative AI systems.15
Intellectual property laws have conventionally granted creators the authority to deter others from using their work and to obtain compensation for their efforts, thereby fostering investment in innovation and the arts. However, since GenAI systems do not necessitate financial or ethical incentives to yield results, this traditional normative framework is called into question, leading to the fundamental debate over whether patent and copyright laws should extend to creations generated by GenAI in the same manner as they do for works created by humans. 16
Due to globalisation of data, copyright regulations have had to adapt, and generative AI is no exception; Article 2 of the Berne Convention restricts copyright protection to works created by humans, which raises challenging questions regarding AI developers, users, or who, should possess rights when AI produces content17. While there is widespread consensus that human creativity is crucial for copyright, the emergence of AI-generated works has complicated this argument, leading scholars such as Kumar (2024)18, Al Da’jeh & Alzubi (2024), and Hutson (2024) to contend that policymakers need to amend prevailing intellectual property laws to elucidate ownership and protection for AI-generated outputs while sustaining incentives for human creators.19
Recent advancements in the interplay of GenAI and intellectual property law emphasise on expounding the conditions under which AI-generated works may be eligible for copyright or patent protection. This includes the enforcement of stringent regulations regarding the use of training data, evidenced by the EU’s AI Act20, which mandates authorization for the extraction of copyrighted materials. Furthermore, there is a strong focus on the indispensability of human authorship or inventorship in countries like the United States and the United Kingdom. Concurrently, regulatory bodies and judicial systems are advocating for increased transparency, the disclosure of training datasets, and a balanced approach to safeguarding the rights of both human creators and innovations driven by AI.21
D. Gendered Implications of GenAI
This study applies Miranda Fricker’s concept of epistemic injustice to examine knowledge-based harms, or epistemic injustices, and to demonstrate how distinct forms of such injustice emerge from the frequently biased nature of GenAI responses towards women, girls, and sexual and gender minorities (SGM). Fricker recognises two main categories of epistemic injustice: testimonial injustice and hermeneutic injustice. 22Testimonial injustice occurs when the testimony of an epistemic subject is unequitably dismissed, rejected, or questioned due to a bias in the interlocutor that undermines the perceived credibility of that speaker. Building upon Fricker’s concepts of testimonial and hermeneutical injustice, this paper argues that gender-biased GenAI systems can undermine the credibility of women and sexual and gender minorities (testimonial injustice) and further obscure their experiences within large scale knowledge repositories (hermeneutical injustice). Through case studies oriented towards work and leadership, they indicate that GenAI is not inherently biased by nature, yet frequently perpetuates structural gender bias, thereby generating unique epistemic harms. Additionally, the author proposes a feminist framework for the analysis and effective mitigation of these harms in future research, which shall be further deliberated upon in this paper.23
The paper examines existing work on gender bias in AI, then uses Fricker’s theory of epistemic injustice and a feminist lens to critically analyze three GenAI case studies, asking:-
• How gender-biased outputs appear?,
• How they can undermine perceptions of women’s competence and reliability? and ;
• How the overarching GenAI ecosystem produces gendered hermeneutical injustice, before concluding with ramifications for future research and policy/practice change?24
Despite the notable potential of generative AI, it has consistently facilitated gender-based harms, facilitating transphobic chatbots, biased and stereotypical portrayals, hyper-sexualized and racist content, as well as abusive applications like deepfakes and AI-generated child sexual abuse material, illustrating that substantial gender and safety risks now accompany the widespread adoption of GenAI.
UNDERSTANDING GENAI AND INTELLECTUAL PROPERTY LAW
The nexus between artificial intelligence (AI) and intellectual property (IP) demonstrates a range of inquiries concerning trademark law, patent law, and copyright law. This study underscores the interplay of generative AI (GenAI) and copyright by providing some technical context and thereafter exploring how various jurisdictions globally address the conflict between GenAI and IP.
A. GenAI: Concept and Functioning
AI systems function autonomously to derive methodologies for accomplishing goals through machine learning or logic-driven techniques, resulting in outputs like content, predictions, or decisions. AI refers to technology that executes tasks usually linked to human intelligence. It typically relies on algorithms to analyze data, recognize patterns, and make decisions or forecasts, and in some cases, it learns and adapts from that experience. 25
An illustrative example in library research is the manner in which online search tools sort results by relevance or recommend resources based on search history, utilizing pattern recognition techniques like word frequency to highlight the most beneficial content.26
The release of ChatGPT by OpenAI in November 2022 marked a major milestone for Generative AI, attracting more than one million users within just five days and sparking considerable public interest. According to a 2023 survey conducted by Deloitte, 61% of computer-based employees in Switzerland were already utilizing ChatGPT or comparable tools on a daily basis. Frequently compared to the introduction of the iPhone, ChatGPT rendered advanced large language models more available to the general public, exhibiting their practical applications and initiating a surge of investment and innovation throughout the sector.27
GenAI boasts an extensive scholarly foundation, beginning with the creation of the first chatbot, ELIZA, in the 1960s. However, contemporary GenAI is distinguished by the utilization of deep learning and neural networks. 28In contrast to conventional rule-based systems that were limited to performing single tasks, today’s models are calibrated using extensive datasets that cover numerous domains, for example, GPT-3 was trained on more than 45 terabytes of compressed text, which imparts an illusion of creativity. This advancement allows tools such as ChatGPT and Google Gemini to engage in adaptive and diverse conversations, as well as to produce not only text but also images, music, and code, far exceeding the scripted and task-specific limitations of traditional chatbots.29
| Year | Development | Significance |
|---|---|---|
| 1957 | Perceptron (Rosenblatt) | Foundation of neural networks |
| 1990s | Statistical Language Models | Early commercial AI (e.g., translation systems) |
| 1997 | LSTM Networks | Enabled sequence learning (text, speech) |
| 2013 | Variational Autoencoders (VAEs) | Early generative models for image creation |
| 2014 | Generative Adversarial Networks (GANs) | Breakthrough in realistic image generation |
| 2017 | Transformer Architecture | Foundation of large language models (LLMs) |
| 2018 | GPT (OpenAI) | First transformer-based generative language model |
| 2019 | GPT-2 | Advanced coherent text generation |
| 2020 | GPT-3 | Large-scale model with human-like language capability |
| 2021 | DALL·E | Text-to-image generation |
| 2021 | GitHub Copilot | AI-assisted coding systems |
| 2022 | Stable Diffusion / MidJourney | Democratization of generative AI tools |
| 2022 | ChatGPT | Mainstream adoption of AI systems |
| 2023 | GPT-4 | Multimodal capabilities (text + image) |
| 2023 | LLaMA 2 (Meta) | Expansion of open-source AI models |
| 2023 | Copyright Litigation (NYT v. OpenAI) | Legal challenges regarding AI training data |
| 2024 | Sora (OpenAI) | Text-to-video generation |
| 2024 | GPT-4o | Real-time multimodal AI interaction |
| 2025 | Advanced Multimodal AI Systems | Wider integration of AI across text, image, video, and real-time applications |
| 2025 | Global AI Regulation Efforts | Increased focus on AI governance frameworks (e.g., EU AI Act implementation) |
| 2026 | Evolving AI Liability Frameworks | Ongoing debates on authorship, accountability, and IP ownership in AI systems |
| 2026 | Human-AI Collaborative Creation | Greater emphasis on co-creation and rethinking ownership in intellectual property law3031 |
Table 1: Evolution of GenAI and Emerging Legal Implications
In order to facilitate a computer to distinguish between a horse and a cow, it is fundamental for human experts to initially instruct the algorithm on the specific features to identify. Without this preliminary instruction, the algorithm may find it challenging to discern the unique traits of each animal. As time progresses, machine learning empowers the system to autonomously detect these patterns and utilize them with new data.32
Notable generative AI technologies include:
• Natural Language Processing (NLP) models: GPT-4, Bard, Claude
• Text-to-image models: DALL·E, MidJourney, Stable Diffusion
• AI-generated music: Jukebox, AIVA
• Code generation models: GitHub Copilot, AlphaCode
GenAI systems strengthen their outputs via iterative feedback until the results accomplish maximum realism. Autoencoders optimize the methods by which these programs store and process data by mitigating noise, with developers enforcing variations during the encoding process to facilitate content creation. Diffusion models, utilized by DALL-E, are trained on images by gradually blurring them and thereafter learning to revert the pixels to their original state, which allows for the creation of new images based on existing ones. However, ChatGPT employs a transformer-based model that assesses context and sequence within the data to make predictions and generate new text.33
B. An Intellectual Property approach to GenAI
As AI continues to evolve in its potential to generate work that rivals human creativity, it raises important questions concerning the assimilation of these technologies into current legal frameworks. The intersection of AI and intellectual property encompasses trademark, patent, and copyright law, with the connection between GenAI and copyright being especially significant, necessitating both a technical comprehension of the functioning of these systems and an analysis of how various nations are addressing this emerging tension.34
GenAI presents two core copyright issues: firstly, in the training phase, these systems utilize extensive datasets that potentially contain copyrighted materials, which raises concerns regarding fair remuneration for creators; secondly, after content is produced, it is ambiguous whether the outputs generated by AI are admissible for copyright protection, considering the persistent deliberations surrounding originality and authorship.35
GenAI is causing significant impact to intellectual property rights in four core areas of law:
• Copyright: The authorship of AI-generated content is unclear, as models may utilize copyrighted materials without authorization, and the classification of outputs as transformative is still a contested issue.
• Patents: Legally, AI cannot be recognized as an inventor, which renders ownership of inventions created by AI in a state of legal uncertainty and poses challenges to conventional standards of inventorship.
• Trademarks: Logos and slogans produced by AI may inadvertently resemble existing trademarks, which could lead to consumer confusion and potential legal conflicts.
• Trade Secrets: The data utilized for training may include confidential business information, necessitating that companies protect AI-generated insights from their competitors.36
C. Copyright Infringement Risks & Suggested Reforms
Prominent GenAI models undergo model training on extensive datasets, which includes books, artwork, and music, frequently without securing consent from copyright owners, thereby generating growing apprehension regarding unauthorized usage.
Consequently, courts must determine whether the outputs produced by AI qualify as transformative fair use or as infringement, taking into account various factors such as the model’s commercial or educational intent, the amount of copyrighted content utilized, and whether the AI-generated results compete with or displace the original creation.
In the case of Google Books v. Authors Guild (2015)37, the court determined that the act of scanning books for the purpose of search indexing was considered fair use, a ruling that could have implications in AI-related legal matters. More recently, in Getty Images v. Stability AI (2023), Getty Images initiated a lawsuit against Stability AI, claiming that it unlawfully used millions of copyrighted images to train its image-generation model. 38Additionally, in Sarah Silverman et al. v. OpenAI and Meta (2023) a collective of authors, including Sarah Silverman, has filed a lawsuit against OpenAI and Meta for the unauthorized use of their works in AI training. These legal actions indicate that, although the legal framework is still evolving, courts are increasingly tasked with clarifying the limits of AI in relation to copyright.39
In order to address the copyright challenges associated with artificial intelligence, a number of legal solutions have been advanced. These solutions include mandatory transparency regarding datasets, which mandates AI companies to reveal the origins of their training data; compensation frameworks established through licensing agreements between AI developers and content creators; as well as a more explicit acknowledgment of human-AI collaboration within copyright law to more precisely delineate authorship in works that involve AI assistance.40
D. Patents and & Suggested Reforms
Contemporary patent law reinforces human inventorship, a principle affirmed in Thaler v. Vidal (2022), where the USPTO held that only individuals can be designated as inventors. 41However, as AI systems become increasingly well-versed in independent innovation, as evidenced by DABUS creating original product designs, DeepMind’s AlphaFold transforming protein-folding research, and IBM’s AI developing new pharmaceutical compounds, urgent legal questions emerge: should AI be acknowledged as an inventor in its own capacity, and do AI-assisted innovations warrant separate patent evaluation within changing legal systems.42
In major countries, patent law uniformly establishes an inventor as a natural person, which presents a significant concern for innovations generated by artificial intelligence. The United States Patent and Trademark Office, the European Patent Office, and the UK Supreme Court have all concluded that AI cannot be recognized as an inventor within existing legal frameworks. Australia has likewise rejected the notion of AI inventorship; however, South Africa represents a notable exception, having become the first nation to issue a patent that names AI as an inventor, indicating that a global consensus on this matter is still not established.43
Litigators often fail to give established answers about AI’s role in invention: how much human input is required for patent validity, whether AI-assisted inventions call for a distinct category, and who ultimately posits ownership, developers, users, or companies. The complexity is highlighted in drug discovery, where an AI may generate a new chemical structure while a human scientist refines and validates it, raising the question of whether the scientist alone qualifies as inventor or whether AI deserves recognition as a co-inventor.44
Research experts have proposed a number of reforms for patents domain: creating separate patent categories that distinguish between inventions generated entirely by AI and those that are AI-assisted; enforcing co-inventorship frameworks that acknowledge AI’s significant contributions; amending the duration of patents to account for AI’s involvement in the innovation process; and instituting disclosure obligations to promote transparency in applications that encompass AI assistance. Collectively, these policy recommendations seek to strike a balance between encouraging innovation and ensuring legal equity and public access to advancements driven by AI.45
E. Trademarks, Trade Secrets & Suggested Reforms
AI-generated logos and slogans may unintentionally replicate pre-existing trademarks, creating difficulties for courts in evaluating distinctiveness and potential consumer ambiguity. Proposed reforms indicate the necessity of human oversight prior to registration, the certification of originality, and the use of AI-driven screening tools to identify similarities with current marks. Regarding liability, legal amendments could make companies responsible for AI-generated infringements and create frameworks aimed at developers whose models contribute to such violations.
Trade secrets pose an analogous issue, as AI systems might reveal proprietary information or allow for the replication of competitors’ technologies. Businesses are encouraged to adopt strong cybersecurity protocols and AI-specific non-disclosure agreements to safeguard confidential data.46
F. A Feminist approach to Intellectual Property and GenAI
International human rights frameworks lay down a foundational basis for the protection of intellectual property, both Article 27(2) of the UDHR and Article 15(1) of the ICESCR recognise the right to safeguard moral and material interests that emerge from an individual’s creative and scientific initiatives. Nevertheless, neither of these instruments provides a definition for “property” or “intellectual property,” which results in the scope of such protections being largely dependent on national legal systems that manifest material variability in their methodologies.47
In addition to these legal deficiencies, deeply embedded gender norms persist in discouraging women from pursuing opportunities in IP-intensive sectors. Those women who do enter these fields often face insufficient support, hostile work environments, wage inequalities, and a lack of mentorship, all of which jointly undermine their confidence and participation. In creative domains such as writing and crafts, IP law aggravates these challenges. For example, handicrafts are frequently excluded from protection on the grounds that they are produced locally or do not comply with legal originality criteria, despite their substantial cultural and economic significance. Collectively, the current IP frameworks are inadequately equipped to tackle the gendered disparities that are integrated in both the legal system and the industries it regulates.48
Women persist to be substantially underrepresented in the intellectual property (IP) framework; only 16% of patent applications filed through the World Intellectual Property Organization (WIPO) emanate from women, emphasising entrenched structural barriers rather than a lack of creativity or capability. 49
• Gender Bias and Discrimination: Prevalent gender norms persist in diverting women from STEM disciplines, which are critical pathways into careers that are intensive in intellectual property. Those who manage to enter these fields experience hostile environments, wage gaps, and insufficient support, all of which as a whole undermine their participation and self-assurance.
• Lack of Representation: Women constitute less than 20% of graduates in STEM and possess only 15% of life science patents, despite securing the majority of PhDs in this area. Further research indicates that inventions by women experience higher rates of rejection, revealing that the concept of inventorship is socially constructed and mirrors existing societal biases.
• Access to Resources and Funding: Financial obstacles disproportionately affect women inventors and entrepreneurs, limiting their ability to file patents or commercialize their inventions. Limited access to loans, venture capital, and grants exacerbates pre-existing inequalities in property and legal rights.
• IP Education and Awareness: A lack of awareness regarding IP rights and processes remains a major obstacle, with the technical complexities of IP law further impeding women’s involvement in the system.
• Work-Family Responsibilities: The unequal distribution of caregiving responsibilities limits the time and resources that women can allocate to IP protection, leading to lower rates of professional growth and commercialization.
• Legal, Procedural, and Networking Barriers: Complicated legal procedures, insufficient representation, and a scarcity of mentorship and networking opportunities further impede women’s capacity to safeguard and leverage their intellectual property, even after they have secured protection.50
Notwithstanding enduring obstacles, women are progressively receiving acknowledgment for their contributions to science, technology, and innovation. Historically marginalized from formal inventorship and authorship, women’s accomplishments are now securing increased recognition, although achieving full and equitable credit persists to be a persistent challenge, especially in male-dominated sectors.
The following instances illustrate the range of women’s creative contributions. Olympia Gonzales, a food technologist, developed new coconut-based products by synthesising traditional cultural knowledge with scientific research, resulting in nutritious solutions rooted in local resources. 51Likewise, Kapinga Mikalu, a Congolese educator, formulated a saliva-based microbe-detection method that utilises both scientific insight and traditional healing practices, motivated by her personal experience of attempting to save her child’s life52. Both examples illustrate how women’s inventions often uniquely assimilate scientific, cultural, and practical knowledge in ways that contest narrow definitions of innovation.
This demonstrates a broader systemic issue: intellectual property frameworks built around the conventional “invention paradigm”, premised in historically masculine norms of rationality and formal scientific methodology, risk marginalising other forms of knowledge creation in which women often excel. Mandating that women conform to this paradigm to achieve recognition may unintentionally diminish alternative epistemologies. A more equitable approach would extend the scope of intellectual property protection to acknowledge multifaceted forms of creativity, while ensuring that the economic rights of all inventors are effectively safeguarded.53
The fundamental principles of intellectual property law are not neutral with respect to gender. Within copyright and patent doctrines lies what feminist legal scholars refer to as the “invention paradigm” a normative structure that portrays the ideal creator as an individual, rational being, implicitly associated with masculinity. Carys Craig contends that the authorship framework in copyright systematically marginalizes collective, communal, and traditionally feminine forms of creativity54. Meanwhile, Ann Bartow outlines that the originality criterion consistently diminishes the value of creative expressions linked to women, such as textile arts and decorative crafts, favoring instead technological and text-based works that have historically been male-dominated.55 In the realm of patent law, the requirement for individual inventorship similarly embodies what Rosemarie Tong describes as a legacy of scientific rationalism that undermines embodied and context-sensitive knowledge. The emergence of GenAI exacerbates these issues: trained on datasets that reflect centuries of gendered exclusion from legal authorship, generative systems do not simply replicate existing biases but amplify them56.
AN EPISTEMIC LENS ON GENDER BIAS AND GENAI
Gender bias refers to the preferential treatment or discrimination directed towards individuals premised on their gender. It is commonly subtle and can appear in extensive forms, affecting individuals in diverse ways according to their gender identity.
There are two categories of gender biases:
• Unconscious bias. Unconscious gender biases are those biases we exhibit without being aware of them. They can mould our decisions and behaviors without our conscious realization.
• Implicit bias. Like unconscious bias, implicit bias covers attitudes or stereotypes that unconsciously affect our perceptions, actions, and choices.57
The analyses of gendered harms in AI has been evaluated through various lenses, including feminist, legal, sociological, and public policy interpretations. Research has emphasised how technology can constrain women’s institutional access, perpetuate damaging stereotypes, and jeopardise bodily autonomy. Feminist epistemology advances additional insight by questioning what constitutes knowledge and who is recognised as a knower, a framework that is critically relevant to GenAI, where the representation of knowledge is pivotal. Although such analyses existed before the onset of modern AI, there is a necessity for their updated application to GenAI. Constructing upon Fricker’s framework, two types of epistemic injustice remain significant: testimonial injustice, which manifests when a speaker’s credibility is wrongfully dismissed due to bias, and hermeneutic injustice, where an individual’s experiences are marginalised within collective interpretive frameworks.58
GenAI’s reliance upon majority data can lead to the systematic exclusion of knowledge from minoritized groups, with research consistently emphasising structural gender biases prevalent in these systems. Instances include text-to-image tools that disproportionately create images of white male professionals, ChatGPT manifests gendered roles when translating from gender-neutral languages, and non-binary individuals being either stereotyped or omitted from representation. For LGBTQIA+ communities, the principal concerns pertain to binary language defaults, misinformation, and the potential for self-censorship in online environments.59
Exacerbating this issue is automation bias, the propensity to excessively depend on automated systems, which makes users more susceptible to accepting biased outputs as objective or credible. Although algorithmic gender bias is not exclusive to GenAI, its function as an epistemic technology that impacts knowledge, decision-making, and content creation intensifies the harms it imposes compared to previous variations of algorithmic bias.60
At present, there exists a minimal convergence of literature exploring the intersection of epistemic injustice and Generative AI (GenAI), notably from a feminist perspective. Kay, Kasirzadeh, and Mohamed introduce the concept of generative algorithmic epistemic injustice, delineating four distinct forms:
• Amplified testimonial injustice – which manifests when pre-existing testimonial injustices are amplified by biases that are structurally embedded during the training and fine-tuning stages of GenAI tools.
• Manipulative testimonial injustice – in which users leverage the functionalities of GenAI to generate content that propagates falsehoods and misinformation, thereby compromising the credibility of others’ testimonies.
• Hermeneutical ignorance – marked by GenAI’s limited sociocultural awareness of human experiences, particularly those of marginalized individuals, leading to the omission and skewing of knowledge belonging to minoritized communities.
• Access injustice – where GenAI systems restrict the ability of minoritized communities to access knowledge.61
A. Case Studies and Causes of Bias
Case Study 1: DALL-E 2 Image Generation
An examination of 15,300 images produced by DALL-E 2 across 153 different professions indicated a greater degree of gender bias compared to Google Images, when evaluated against U.S. census data:
• Men were generated significantly more than women (61.6% vs 38.4%)
• Women were more frequently depicted smiling and looking downward, particularly in female-dominated roles
• Women were underrepresented in male-dominated occupations and overrepresented in female-dominated ones62
Case Study 2: Leadership Narratives on Wordplay.ai
An examination of 50 leadership narratives totalling over 40,000 words uncovered nuanced yet persistent gender bias:
• Women were approximately twice as likely to be portrayed as poor leaders compared to men
• Female bad leadership was framed around incompetence; male bad leadership around self-interest
• Positive portrayals of women emphasised endurance or supporting men, while men were depicted as bold and influential
• Women were evaluated on likeability; men on authority.63
Both case studies demonstrate that gender bias in GenAI is not always overt but manifests through patterns that systematically undermine perceptions of women’s competence and credibility.
Case Study 3: Cultural and Gender Bias in GPT-2 and Llama2 (UNESCO)
Prompting both models to generate occupations for individuals from different cultural backgrounds revealed compounding gender and cultural stereotyping:
• Occupations generated for British men were varied; those for British women were narrower and included stereotypically controversial roles in roughly 30% of cases
• Zulu men received varied but stereotype-reliant responses; Zulu women were overwhelmingly assigned domestic service roles, appearing in approximately 20% of outputs64
Cross-Case Findings:
In the analysis of the three case studies, it was highlighted that gender bias manifested more in presentation than in representation:
• Women were often depicted in a negative light, characterized as ineffective leaders, placed in subordinate positions, or exhibiting body language that is downward-facing.
• Women were regularly pigeonholed into traditionally feminized roles and examined based on feminized traits.
• While some biases were explicit, many were more nuanced, exemplified by gendered variations in the framing of leadership, rendering them more challenging for individuals to identify and likely more pervasive than is currently recognized.65
B. Gender Bias in GenAI and Its Impact on Perceptions of Women’s Competence and Credibility
A study conducted by Grenny and Maxfield demonstrates an empirical gender bias in assessments of workplace competency. When women are considered forceful or assertive, their perceived competence decreases by 35%, and their estimated value diminishes by $15,088, a pronounced difference to the 22% diminution in competence and a $6,547 reduction in worth that men face under analogous assessments. This difference is especially material in leadership roles, where assertiveness is a key attribute, yet women are disproportionately penalized compared to their male peers.66
Gender-biased outputs from GenAI pose a risk of amplifying what Medina refers to as “meta-blindness”67 a phenomenon where individuals utilising epistemically biased resources become unconscious of their own blind spots, often overlooking viewpoints that contest prevailing narratives. This threat is intensified by the multifaceted nature of GenAI’s gender stereotyping, which frequently generates biased content in ways that seem familiar and objective, complicating users’ ability to recognize it. Given the persistent societal attitudes towards gender inequality and an increasing backlash against DEI initiatives, such biases may remain unnoticed or even intentionally reinforced.
As trends suggest that GenAI content will surpass human-generated content online by 2026, these dynamics are expected to intensify. Medina’s suggested remedy, epistemic friction, which entails the practice of holding and comparing conflicting standpoints, seems challenging to acknowledge in a context shaped by majoritarian data and low levels of media literacy. Research indicates that less than half of the adults surveyed could confidently determine misinformation online, emphasising this vulnerability. Ultimately, the epistemic injustices embedded in GenAI threaten to taint the wider epistemic landscape, diminishing trust between speakers and audiences and sustaining systemic patterns of credibility erosion and misinformation.68
Addressing these issues demands for a collaborative framework among governments, the private sector, and civil society, a concept referred to by scholars as feminist technology diplomacy. On a technical level, Ferrara (2024) highlights three essential strategies:
• First, screening training data for bias prior to model development;
• Second, selecting models that are inherently designed with fairness in mind; and
• Third, modifying outputs to ensure equitable outcomes across diverse demographic groups. 69
No single strategy is adequate on its own, and a convergence of all three is generally considered the most effective approach. Some developers have already sought progress in this area, for example, OpenAI has revised DALL-E 2 to more accurately represent global diversity, yet the manner in which women and minorities are characterised, rather than merely represented, continues to be a largely unaddressed issue that requires further investigation.
However, technical solutions alone are insufficient. It is challenging to effectively mitigate hermeneutical injustice without the inclusion of the communities most affected by these, especially women of colour and gender minorities, in the development process itself. While Fricker’s focus on fostering principles such as open-mindedness and critical reflection serves as a valuable starting point, critics contend that individual transformation cannot substitute the need for systemic reform. Addressing bias in GenAI ultimately necessitates both: users who can critically interact with AI-generated content, and structural modifications to the systems that create it, starting with AI literacy education that empowers individuals to identify and question bias in digital environments.70
ARTIFICIAL PROMISES OF GOVERNANCE?
GenAI presents complex governance challenges that necessitate urgent and coordinated policy responses, with data governance and intellectual property emerging as core concerns. Training these models relies on large-scale quantities of internet-sourced data, often without the consent of rights holders, initiating high-profile legal disputes and raising broader questions around copyright, fairness, and creator compensation. Compounding this, existing legal frameworks remain ill-equipped to resolve uncertainties around the copyright status of AI-generated content, human-AI co-created works, and derivative outputs.71
Beyond the issues surrounding intellectual property, GenAI amplifies inherent biases on a large scale, thereby causing discriminatory results based on race, gender, and other protected attributes. Furthermore, its inconsistent application threatens to widen the digital disparity. The ability of this technology to create realistic synthetic media also assists the swift dissemination of misinformation and disinformation, which compromises social unity and public confidence, issues that necessitate both technical interventions and extensive stakeholder collaboration to augment digital resilience.
Moreover, these potential harms are exacerbated by the fact that GenAI significantly reduces the barriers to fraud and cybercrime. Malicious users can take advantage of deepfakes and synthetic media to damage reputations, enable phishing and extortion, create non-consensual intimate images, and manipulate public sentiment on a large scale. Furthermore, this technology reinforces the creation of harmful code and autonomous cyberattacks, presenting serious risks to personal safety, national security, and the global cyber security framework.72
C. The Need for Reformation
A notable governance gap exists in the limited public involvement in the development and implementation of GenAI. Although its societal implications are expansive, decision-making processes tend to be predominantly technocratic, which poses a risk of concentrating exclusively on risk management without effectively considering broader social values or the perspectives of those most impacted. The public sector, which is pivotal for effective governance, faces its own set of challenges, such as a shortage of skills, information asymmetry, the swift pace of technological advancements that surpass legislative measures, and the complexities of consistently regulating a general-purpose technology across various sectors, often leading to regulations that are reactive, ambiguous, and insufficiently detailed.73
On an international scale, the intrinsically global character of AI development entails coordinated efforts in standard-setting, sharing best practices, and resolving transboundary risks, especially in relation to autonomous weapons and the competitive environment of the US-China AI rivalry. Tackling these concerns requires a transition from merely reactive governance solutions to a more holistic, systemic strategy, one that unites policymakers, industry stakeholders, researchers, and the public to facilitate trust in the technology and ensure it benefits the broader community.74
D. Challenges and Proposed Solutions
1. Privacy and Personal Data Risks in GenAI
Issue: “By 2027, more than 40% of AI-related data breaches will be caused by the improper use GenAI across borders,” according to Gartner, Inc. 75GenAI poses significant privacy and personal data risks across its system lifecycle. Training datasets commonly incorporate personal information extracted from public sources, including names, addresses, and behavioral data, while user interactions further disclose sensitive details about individuals’ thoughts, experiences, and identities. Unlike traditional surveillance, GenAI pervades deeply into personal spaces and inner lives, with conversations feeding back into training data. Furthermore, the distinction between privacy and personal data protection is often conflated, despite these being legally and conceptually distinct, a confusion that weakens regulatory responses across jurisdictions.
Proposed Solution: Addressing these risks mandates clear legal frameworks that separately and effectively govern privacy and personal data protection, as manifested in instruments like the EU’s GDPR and China’s PIPL. Regulators must ensure that data collection practices across all phases of AI development, from pre-training to deployment, are transparent, consent-based, and subject to meaningful oversight. Developers should lessen the use of identifiable personal data in training, enforce strong data security measures, and provide users with genuine control over how their information is used. A risk-based governance approach that accounts for the unique and dynamic nature of GenAI’s data practices is essential to safeguarding individual rights in the digital ecosystem.
2. Legal Complexities and Copyright
Issue: Contemporary copyright systems designed around human authorship as recognised by the Berne Convention, are inherently inadequate for dealing with AI generated content, data training practices, and ownership concerns. Conventional legal mechanisms for copyright, patent, and trademark protection were developed with human creators in mind, resulting in considerable ambiguity regarding the protection, licensing, and enforcement of rights in works produced by, or with the assistance of, AI. Prominent legal disputes have exacerbated these issues: cases involving OpenAI, Stability AI, and the New York Times challenge the limits of fair use, while in 2025, three federal district court rulings began to outline the parameters of what constitutes fair use in relation to AI training on copyrighted materials, with courts arriving at varying conclusions. Jurisdictional discrepancies further exacerbate the situation, as the US, EU, UK, and China each pursue different strategies regarding AI and copyright.76
Solutions: Emerging Legislative and Policy Responses:
• United States: In January 2025, the US Copyright Office upheld that material generated entirely by AI is not eligible for copyright protection, while works that involve human assistance may qualify based on the extent of human contribution.
• European Union: The EU AI Act, effective since August 2024, directs that AI providers reveal details about their training data, comply with exceptions for text and data mining, and label outputs generated by AI.
• United Kingdom: An amendment proposed to the Data Bill would enable the government to mandate that AI companies report their utilization of copyrighted materials, thereby promoting transparency in licensing.
• India: The AI Governance Guidelines of India, set for 2025, foresee amendments to the Copyright Act to tackle issues related to AI-generated content, highlighting a philosophy of “innovation over restraint.”77
• Australia: In October 2025, the Attorney General dismissed a proposed exception for text and data mining, instead pledging to establish a framework that ensures creators receive “fair remuneration” under equitable terms of use.78
• Global: Coordinated international efforts through WIPO and cross-border licensing frameworks are crucial for standardization and to avert regulatory fragmentation.
Expanding on the present law: The guidelines released by the US Copyright Office in March 2023 elucidates that the ability to copyright an AI-assisted work is contingent upon a fundamental question: how much of the work was genuinely created by a human. When the AI is responsible for making the primary creative decisions, including the choice of words, the structure, and the style, the resulting work does not qualify for copyright protection. The guidance states clearly: if an individual instructs a text generator to compose a poem regarding copyright law in the manner of William Shakespeare, it is the AI that decides the rhyming scheme, the wording, and the overall structure of the piece. Since the creative decisions are made by the machine rather than the individual, the output is not eligible for copyright registration.79
However, the scenario shifts when a human takes a more interactive role. If an individual significantly edits, organizes, or modifies AI-generated content in an approach that demonstrates their own creative judgment, that human input may be eligible for copyright protection. Nevertheless, the copyright will only extend to the elements that the human has actually created, the segments generated by the AI will remain unprotected, irrespective of the appearance of the final work.
3. GenAI and Intellectual Property Infringement
Issue: Generative AI tools are developed using extensive amounts of IP-protected content, thereby contributing to persistent legal conflicts globally regarding whether scraping, model training, and AI-generated outputs amount to violations of copyright, trademark, and other intellectual property laws. The legal ambiguity is notable, with potential outcomes differing significantly across various jurisdictions. Crucially, liability does not rest exclusively with developers; in numerous countries, infringement can occur without intent or knowledge, which means that both businesses and end users may face similar risks. Practical Approaches to Managing: IP Infringement Risks in GenAI While legal frameworks are perpetually developing, organizations must not wait for judicial clarity, proactive risk management is vital.
The following measures illustrate suggested reforms:
• Select IP-compliant tools: Emphasize GenAI platforms that are trained on licensed, public domain, or proprietary datasets, and assess whether providers offer indemnities against copyright infringement, including the extent and conditions of that protection.
• Thoroughly vet training datasets: Organizations should scrutinise datasets for open-source obligations, confirm IP ownership, and ensure license coverage for AI training purposes prior to fine-tuning or deploying models. 80
• Document human contributions: It is fundamental to clearly record human involvement in AI-assisted projects and establish agreements with model providers and developers to outline ownership and IP rights, thereby bolstering any future copyright claims.81
• Screen outputs prior to use: Leveraging plagiarism detection software, reverse image search tools, and code license scanners can aid in identifying obvious infringements, while human editing and rewriting can minimize similarity to any single source.82
• Implement technical safeguards: Innovative frameworks such as watermarking, fingerprinting, and machine unlearning present new strategies to protect copyrights within the AI domain and mitigate infringement at the generation stage.
• Train staff and enforce policies: Organizations should proactively establish appropriate policies and provide training to employees to ensure the safe and fair use of generative AI, including guidance against prompts that reference third-party trademarks, copyrighted works, or named artists.
• Monitor regulatory developments: Transparency requirements regarding training data disclosure are currently under consideration across various jurisdictions, and compliance strategies should be adapted accordingly as binding regulations are formulated.83
4. Intersectional lens on GenAI and Gender Bias
Building on the structural barriers outlined in the section of epistemic lens on GenAI and Gender bias, the integration of GenAI into IP ecosystems risks compounding existing gender inequities, embedding and scaling the same biases that have historically excluded women from invention, authorship, and legal recognition.
Suggested reforms include:
• Targeted IP education: Enforce accessible IP literacy programs aimed at women and girls, especially in STEM and creative sectors, to clarify filing procedures and emphasize the economic benefits of IP protection.
• Inclusive funding mechanisms: Create gender-responsive grants, subsidized patent filing initiatives, and specific venture capital opportunities for women inventors and entrepreneurs.
• Mentoring and networking infrastructure: Broaden initiatives like ChIPs (the networking organization for women in IP law founded in Silicon Valley) and establish formal sponsorship programs that connect women with experienced IP professionals.
• Reform IP originality standards: Expand the legal definition of protectable works to incorporate traditionally female-associated forms of creativity, such as handicrafts, textile arts, and domestic innovations that are currently excluded due to originality criteria.
• Feminist AI governance frameworks: Initiatives like the FA+IR Feminist AI Research Network are creating practical prototypes such as SOF+IA to directly combat technology-facilitated gender-based violence, illustrating that feminist involvement in AI design is both feasible and essential.84
• Regional and grassroots networks: Networks in Latin America, such as Data against Feminicide, demonstrate how community-driven feminist AI frameworks can both resist and tackle gender-based harms that are caused or exacerbated by AI systems.85
• Global South perspectives in digital governance: The Charter of Feminist Demands from the Global South advocates for digital governance frameworks that prioritize women’s human rights, highlighting the importance of security, transparency, and accountability, principles that should be integrated into any AI and intellectual property reform agenda.86
• Bias auditing of GenAI tools used in IP processes: When AI is utilized in patent screening, trademark evaluation, or legal research, regular bias audits should be required to secure that outputs do not systematically disadvantage women applicants.
• Diverse development teams: GenAI tools that influence IP decisions must be created and evaluated by gender-diverse teams to circumvent the incorporation of male-centric biases into automated systems.
• Transparency in AI training data: Mandate the disclosure of whether the datasets employed to train IP-related GenAI tools adequately mirror women inventors, creators, and legal professionals, and require remedial measures when they do not. 87
• Policy alignment: WIPO and national IP offices should standardize the collection and reporting of gender-disaggregated data, facilitating evidence-based reforms and accountability across different jurisdictions.88
CONCLUSION
The convergence of Generative AI, intellectual property law, and epistemic injustice exposes a profound structural tension within contemporary technological governance. Rather than functioning as neutral tools of innovation, AI systems reflect and reproduce the historical inequalities embedded within the legal and epistemic systems from which they emerge, thereby posing a considerable risk of exacerbating the marginalization of women and marginalised communities in both creative and legal domains. Current intellectual property rights frameworks, although neutral in their design, have repeatedly failed to acknowledge or reward the creative contributions of those historically excluded from formal authorship and ownership systems. GenAI exacerbates these shortcomings by commodifying collective knowledge, expanding the reproduction of gendered stereotypes, and perpetuating epistemic injustices that marginalize the lived experiences of women and LGBTQIA+ individuals within prevailing knowledge frameworks.89
This conflict is anchored in a more profound doctrinal issue. The foundational principles of copyright and patent law, which focus on the image of the solitary, rational, and implicitly masculine creator, directly contradict feminist perspectives that view knowledge as collective, relational, and intersectional. As AI continues to distort the lines of authorship and ownership, these tensions become increasingly significant. Challenges relating to consent, compensation, and creative attribution in AI training and outputs remain largely unaddressed, necessitating legal frameworks that transcend mere formal neutrality in favor of substantive equity.
Nevertheless, initiatives led by feminists present significant justifications for hope. Organizations like FA+IR, grassroots efforts such as Data Against Feminicide, and policy frameworks like the Charter of Feminist Demands from the Global South emphasize that it is both theoretically sound and practically feasible to achieve inclusive, ethical, and accountable governance of AI. These initiatives, embedded in intersectional design, ethical data practices, and open-source principles, offer a pivotal framework for rethinking the governance of AI and intellectual property in ways that prioritize equity, collective ownership, and the viewpoints of those most negatively impacted by existing systems.90
*****
Footnotes
1. Foundation for Media Alternatives (FMA), Algorithms of Abuse: Addressing Online Gender-Based Violence in the Age of Artificial Intelligence, Foundation for Media Alternatives (Sept. 26, 2024), https://fma.ph/algorithms-of-abuse-addressing-online-gender-based-violence-in-the-age-of-artificial-intelligence/.
2. A Brief History of Generative AI, Dataversity, https://www.dataversity.net/articles/a-brief-history-of-generative-ai/ (last visited January 3, 2026).
3. The Rise of Generative AI, https://trendsresearch.org/insight/the-rise-of-generative-ai/ (last visited March 11, 2026).
4. Manu Jain, How to Build a Generative AI Model in 2025?, ScaleupAlly (Jan. 28, 2024), https://scaleupally.io/blog/how-to-build-generative-ai/.
5. Generative AI Applications, GeeksforGeeks (19:59:21+00:00), https://www.geeksforgeeks.org/artificial-intelligence/generative-ai-applications/.
6. AI Market Projected to Hit $4.8 Trillion by 2033, Emerging as Dominant Frontier Technology | UN Trade and Development (UNCTAD), (Apr. 7, 2025), https://unctad.org/news/ai-market-projected-hit-48-trillion-2033-emerging-dominant-frontier-technology.
7. AI Investment Forecast to Approach $200 Billion Globally by 2025 | Goldman Sachs, https://www.goldmansachs.com/insights/articles/ai-investment-forecast-to-approach-200-billion-globally-by-2025 (last visited May 1, 2026).
8. Albert Bigelow Paine, Mark Twain, a Biography: The Personal and Literary Life of Samuel Langhorne Clemens (1912).
9. (PDF) Creativity and Arts in Digital Social Innovation, ResearchGate, https://www.researchgate.net/publication/373159149_Creativity_and_Arts_in_Digital_Social_Innovation (last visited February 12, 2026).
10. Where Good Ideas Come From: The Natural History of Innovation by Steven Johnson | Goodreads, https://www.goodreads.com/book/show/8034188-where-good-ideas-come-from (last visited April 25, 2026).
11. Artificial Imagination: The Rise of Generative AI | Berkeley Exec Ed, https://executive.berkeley.edu/thought-leadership/blog/artificial-imagination-rise-generative-ai (last visited January 2, 2026).
12. Baptiste Caramiaux et al., Generative AI and Creative Work: Narratives, Values, and Impacts (2025).
13. The State of AI: Global Survey | McKinsey, https://www.mckinsey.com/capabilities/quantumblack/our-insights/the-state-of-ai-how-organizations-are-rewiring-to-capture-value (last visited January 23, 2026).
14. Lydia Boussour Senior Economist https://www.ey.com/en_gl/people/lydia-boussour EY-Parthenon, Ernst & Young, Harness the Productivity Potential of GenAI, (last visited March 14, 2026).
15. The Interaction between Intellectual Property Laws and AI: Opportunities and Challenges, https://www.nortonrosefulbright.com/en/knowledge/publications/c6d47e6f/the-interaction-between-intellectual-property-laws-and-ai-opportunities-and-challenges, https://www.nortonrosefulbright.com/en/knowledge/publications/c6d47e6f/the-interaction-between-intellectual-property-laws-and-ai-opportunities-and-challenges (last visited February 22, 2026).
16. Ethical Implications of Using Artificial Intelligence in Intellectual Property Creation: Authorship, Ownership and Responsibility Issues | Afuwape | Journal of Digital Technologies and Law, https://www.lawjournal.digital/jour/article/view/590?locale=en_US (last visited January 23, 2026).
17. WIPO Lex, https://www.wipo.int/wipolex/en/text/283698 (last visited March 2, 2026).
18. Kumar, S. (2024) Inductive and Deductive Approaches to Qualitative Research. International Journal of Multidisciplinary Educational Research, 13, 58-63. – References – Scientific Research Publishing, https://www.scirp.org/reference/referencespapers?referenceid=3987096 (last visited March 2, 2026).
19. Afuwape, K. (2025). Ethical Implications of Using Artificial Intelligence in Intellectual Property Creation: Authorship, Ownership and Responsibility Issue. Journal of Digital Technologies and Law, 3(4), 677–704. https://doi.org/10.21202/jdtl.2025.27
20. EU Artificial Intelligence Act | Up-to-Date Developments and Analyses of the EU AI Act, https://artificialintelligenceact.eu/ (last visited March 22, 2026).
21. Adil S. Al-Busaidi et al., Redefining Boundaries in Innovation and Knowledge Domains: Investigating the Impact of Generative Artificial Intelligence on Copyright and Intellectual Property Rights, 9 Journal of Innovation & Knowledge 100630 (2024).
22. Evolving Concepts of Epistemic Injustice, in The Routledge Handbook of Epistemic Injustice 53 (2017), https://www.taylorfrancis.com/chapters/edit/10.4324/9781315212043-5/evolving-concepts-epistemic-injustice-miranda-fricker.
23. Isobel Barry & Elise Stephenson, The Gendered, Epistemic Injustices of Generative AI, 40 Australian Feminist Studies 1 (2025).
24. Jackie Kay, Atoosa Kasirzadeh & Shakir Mohamed, Epistemic Injustice in Generative AI (Aug. 21, 2024), http://arxiv.org/abs/2408.11441.
25. What Is Artificial Intelligence (AI)?, Palo Alto Networks, https://www.paloaltonetworks.com/cyberpedia/artificial-intelligence-ai (last visited May 1, 2026).
26. Angela Young, Library Guides and Databases: Generative AI and Library Skills: Generative AI and Library Skills, https://library-guides.ucl.ac.uk/generative-ai/home (last visited March 11, 2026).
27. Generative AI’s Fast and Furious Entry into Switzerland | Deloitte Switzerland, https://www.deloitte.com/ch/en/Industries/technology/research/generative-ai.html (last visited January 2, 2026).
28. Generative Artificial Intelligence: A Historical Perspective | National Science Review | Oxford Academic, https://academic.oup.com/nsr/article/12/5/nwaf050/8029900 (last visited April 12, 2026).
29. Hooman H. Rashidi et al., Introduction to Artificial Intelligence and Machine Learning in Pathology and Medicine: Generative and Nongenerative Artificial Intelligence Basics, 38 Modern Pathology 100688 (2025).
30. Generative Artificial Intelligence: A Historical Perspective | National Science Review | Oxford Academic, https://academic.oup.com/nsr/article/12/5/nwaf050/8029900 (last visited February 14, 2026).
31. Entries from 2025–2026 reflect emerging trends and ongoing legislative developments rather than settled law.
32. AI vs. Generative AI: The Differences Explained | Coursera, https://www.coursera.org/articles/ai-vs-generative-ai (last visited February 14, 2026).
33. A Systematic Review of Generative AI: Importance of Industry and Startup-Centered Perspectives, Agentic AI, Ethical Considerations & Challenges, and Future Directions | Artificial Intelligence Review | Springer Nature Link, https://link.springer.com/article/10.1007/s10462-025-11435-z (last visited March 11, 2026).
34. Ties between Generative Artificial Intelligence and Intellectual Property Rights – Diplo, (June 26, 2024), https://www.diplomacy.edu/blog/ties-between-generative-artificial-intelligence-and-intellectual-property-rights/.
35. The Law and Economics of Generative AI and Copyright: A Primer to Core Challenges for Our Digital Future – PMC, https://pmc.ncbi.nlm.nih.gov/articles/PMC12658474/ (last visited March 11, 2026).
36. Daniel Chibuike Okoro, An Intellectual Property Approach to the Use of Generative AI, 9 (2025).
37. Authors Guild v. Google, Inc., No. 13-4829 (2d Cir. 2015) :: Justia, https://law.justia.com/cases/federal/appellate-courts/ca2/13-4829/13-4829-2015-10-16.html (last visited March 2, 2026).
38. Getty Images v Stability AI English High Court Rejects Secondary Copyright Claim, https://www.lw.com/en/insights/getty-images-v-stability-ai-english-high-court-rejects-secondary-copyright-claim (last visited March 1, 2026).
39. Sarah Silverman Is Suing OpenAI and Meta for Copyright Infringement | The Verge, https://www.theverge.com/2023/7/9/23788741/sarah-silverman-openai-meta-chatgpt-llama-copyright-infringement-chatbots-artificial-intelligence-ai (last visited March 1, 2026).
40. Copyright and AI Training Data—Transparency to the Rescue? | Journal of Intellectual Property Law & Practice | Oxford Academic, https://academic.oup.com/jiplp/article/20/3/182/7922541 (last visited March 2, 2026).
41. The Legal World’s Dilemma with AI Inventorship, https://www.entefylabs.ai/insights/the-legal-worlds-dilemma-with-ai-inventorship (last visited March 2, 2026).
42. Review of AlphaFold 3: Transformative Advances in Drug Design and Therapeutics – PMC, https://pmc.ncbi.nlm.nih.gov/articles/PMC11292590/ (last visited March 2, 2026).
43. Revised USPTO Guidance on Inventorship for AI-Assisted Inventions: A pro-Innovation Pivot Away from Pannu Factors | Journal of Intellectual Property Law & Practice | Oxford Academic, https://academic.oup.com/jiplp/advance-article/doi/10.1093/jiplp/jpag021/8528856 (last visited March 2, 2026).
44. AI and Inventorship Guidance: Incentivizing Human Ingenuity and Investment in AI-Assisted Inventions | USPTO, https://www.uspto.gov/blog/ai-and-inventorship-guidance-incentivizing (last visited March 2, 2026).
45. AI-Generated Inventions: New Questions for Patent Regimes, https://www.snrlaw.in/ai-generated-inventions-new-questions-for-patent-regimes/ (last visited March 2, 2026).
46. AI-Generated Logos and Trademark Eligibility: Navigating Intellectual Property Law in the Age of Artificial Intelligence by Anshuman Singh :: SSRN, https://papers.ssrn.com/sol3/papers.cfm?abstract_id=6564883 (last visited March 2, 2026).
47. Peter Yu, The Anatomy of the Human Rights Framework for Intellectual Property, 69 SMU L. REV. 37 (2016)
48. Challenges Faced by Women in Protecting Intellectual Property Rights | International Journal of Law Management & Humanities, https://ijlmh.com/paper/challenges-faced-by-women-in-protecting-intellectual-property-rights/ (last visited March 1, 2026).
49. Gender Equality and Intellectual Property, women, https://www.wipo.int/en/web/women (last visited March 4, 2026).
50. Azhar T. Rahma et al., Factors Affecting Women Scientists’ Retention and Progress in STEM Fields in the UAE: A Cross-Sectional Study, 13 F1000Res 1539 (2026).
51. M. Akshaya et al., Ethnobotany of Coconut: A Review on Its Traditional, Cultural and Economic Significance, 12 Plant Science Today (2025), https://horizonepublishing.com/index.php/PST/article/view/11221.
52. Diaz, John & Gusto, Cody & McCoy, Kaci & Silvert, Colby & Bala, Joseph & Atibu, Joseph & Tshefu, Antoinette & Mwandagalirwa, Melchior & Dinglasan, Rhoel. (2023). A mixed methods study assessing the adoption potential of a saliva-based malaria rapid test in the Democratic Republic of Congo. Malaria Journal. 22. 10.1186/s12936-023-04599-y.
53. Aneesa Nahan, A FEMINIST PERSPECTIVE OF WOMEN’S INTELLECTUAL PROPERTY RIGHTS AND DRAWBACKS IN COPYRIGHT LAWS, International Journal for Legal Research and Analysis (2023), https://www.ijlra.com/public/details/a-feminist-perspective-of-women’s-intellectual-property-rights-and-drawbacks-in-copyright-laws-by-aneesa-nahan.
54. Carys J. Craig, Copyright and Gender: Feminist Philosophies and the Politics of Proof (Apr. 18, 2024), https://papers.ssrn.com/abstract=4905148.
55. Cases | Springer Nature Link, https://link.springer.com/chapter/10.1007/978-981-97-6095-4_3 (last visited May1, 2026).
56. Misinformation Studies and Higher Education in the Postdigital Era: Beyond Fake News, https://www.researchgate.net/publication/388846947_Misinformation_Studies_and_Higher_Education_in_the_Postdigital_Era_Beyond_Fake_News (last visited May 1, 2026).
57. Understanding Gender Bias: Key Issues and Strategies for Change, (Sept. 26, 2024), https://www.imd.org/blog/management/gender-bias/.
58. Seyyed Kazem Banihashem et al., The Intersection of Epistemic Beliefs and Gender in Argumentation Performance, 61 Innovations in Education and Teaching International 716 (2024).
59. Jerlyn Q.H. Ho et al., Gender Biases within Artificial Intelligence and ChatGPT: Evidence, Sources of Biases and Solutions, 4 Computers in Human Behavior: Artificial Humans 100145 (2025).
60. Kate Goddard, Abdul Roudsari & Jeremy C Wyatt, Automation Bias: A Systematic Review of Frequency, Effect Mediators, and Mitigators, 19 J Am Med Inform Assoc 121 (2012).
61. Jackie Kay, Atoosa Kasirzadeh & Shakir Mohamed, Epistemic Injustice in Generative AI (2024), https://arxiv.org/abs/2408.11441.
62. Luhang Sun et al., Smiling Women Pitching down: Auditing Representational and Presentational Gender Biases in Image-Generative AI, 29 Journal of Computer-Mediated Communication zmad045 (2023).
63. Toby Newstead, Bronwyn Eager & Suze Wilson, How AI Can Perpetuate – Or Help Mitigate – Gender Bias in Leadership, 52 Organizational Dynamics 100998 (2023).
64. https://unesdoc.unesco.org/ark:/48223/pf0000388971 (last visited January 2, 2026).
65. Barry, I. and Stephenson, E. (2025) ‘The Gendered, Epistemic Injustices of Generative AI’, Australian Feminist Studies, 40(123), pp. 1–21. doi: 10.1080/08164649.2025.2480927.
66. Melissa Miller Young, Gender Bias Still At Play In The Workplace, Dr. Nancy (Aug. 31, 2015), https://www.drnancyoreilly.com/gender-bias-still-at-play-in-the-workplace/.
67. Roberto Frega, José Medina, The Epistemology of Resistance: Oxford, Oxford University Press, 2012, 352 Pages, V ejpap (2013), https://journals.openedition.org/ejpap/678.
68. Liming Pan et al., Complexity of Social Media in the Era of Generative AI, 12 National Science Review nwae323 (2024).
69. Emilio Ferrara, Fairness and Bias in Artificial Intelligence: A Brief Survey of Sources, Impacts, and Mitigation Strategies, 6 Sci 3 (2023).
70. OpenAI: Building the “Everything Platform” in AI – Leonis Capital, https://www.leoniscap.com/research/op enai-building-the-everything-platform-in-ai (last visited February 11, 2026).
71. Xiongbiao Ye et al., Privacy and Personal Data Risk Governance for Generative Artificial Intelligence: A Chinese Perspective, 48 Telecommunications Policy 102851 (2024).
72. Araz Taeihagh, Governance of Generative AI, 44 Policy and Society 1 (2025).
73. Zifan Chen, Ying Wang & Xiuyuan Zhao, Responsible Generative AI: Governance Challenges and Solutions in Enterprise Data Clouds, 18 JCEIM 59 (2025).
74. Laure de Roucy-Rochegonde, “Artificial Promises or Real Regulation? Inventing Global AI Governance”, Ifri Studies, Ifri, February 2025.
75. Gartner Predicts 40% of AI Data Breaches Will Arise from Cross-Border GenAI Misuse by 2027, https://www.gartner.com/en/newsroom/press-releases/2025-02-17-gartner-predicts-forty-percent-of-ai-data-breaches-will-arise-from-cross-border-genai-misuse-by-2027 (last visited March 5, 2026).
76. Copyright law in 2025: Courts begin to draw lines around AI training, piracy, and market harm | Reuters, https://www.reuters.com/legal/legalindustry/copyright-law-2025-courts-begin-draw-lines-around-ai-training-piracy-market-harm–pracin-2026-03-16/
77. India AI Governance Guidelines, https://www.pib.gov.in/www.pib.gov.in/Pressreleaseshare.aspx?PRID =2228315 (last visited March 5, 2026).
78. Albanese government to ensure Australia is prepared for future copyright challenges emerging from ai | our ministers – attorney-general’s portfolio. https://ministers.ag.gov.au/media-centre/albanese-government-ensure-australia-prepared-future-copyright-challenges-emerging-ai-26-10-
79. Copyright and Artificial Intelligence | U.S. Copyright Office, https://www.copyright.gov/ai/ (last visited March 5, 2026).
80. Artificial Intelligence and Open Source Data and Software: Contrasting Perspectives, Legal Risks, and Observations, Crowell & Moring – Artificial Intelligence and Open Source Data and Software: Contrasting Perspectives, Legal Risks, and Observations, https://www.crowell.com/en/insights/client-alerts/artificial-intelligence-and-open-source-data-and-software-contrasting-perspectives-legal-risks-and-observations (last visited March 5, 2026).
81. Admin Administrator, AI-Assisted Creative Works: How to Prove Human Creativity, Copyright Ownership, Asia IP, https://asiaiplaw.com/section/in-depth/ai-assisted-creative-works-how-to-prove-human-creativity-copyright-ownership (last visited April 25, 2026).
82. Martensen IP, Martensen IP Offers Critical Guidance on Understanding the Essentials of AI Risk and Intellectual Property Compliance, GlobeNewswire News Room (Dec. 1, 2025), https://www.globenewswire.com/news-release/2025/12/01/3197492/0/en/Martensen-IP-Offers-Critical-Guidance-on-Understanding-the-Essentials-of-AI-Risk-and-Intellectual-Property-Compliance.html.
83. Generative Artificial Intelligence Use Policy or Guidelines for Employees, (Mar. 6, 2024), https://www.ue.org/risk-management/the-workplace/generative-ai-use-policy-or-guidelines-for-employees/.
84. Feminist AI Research Network: Combatting Gender-Based Violence with Artificial Intelligence Innovations | IDRC – International Development Research Centre, https://idrc-crdi.ca/en/research-in-action/feminist-ai-research-network-combatting-gender-based-violence-artificial (last visited May 5, 2026).
85. Paola Ricaurte Quijano et al., Feminist AI for/by the Majority World: Feminist AI Research Network, Latin American and Caribbean Hub, in Oxford Intersections: AI in Society (Philipp Hacker ed.), https://doi.org/10.1093/9780198945215.003.0085 (last visited May 1, 2026).
86. Charter of feminist demands from the Global South: A Global Digital Compact for gender equality | it for change. https://itforchange.net/charter-of-feminist-demands-from-global-south-a-global-digital-compact-for-gender-equality (last visited 01 May 2026).
87. Bias of AI-Generated Content: An Examination of News Produced by Large Language Models | Request PDF, ResearchGate, https://www.researchgate.net/publication/374714914_Bias_of_AI-Generated_Content_An_Exa mination_of_News_Produced_by_Large_Language_Models (last visited March 25, 2026).
88. Innovation Gender Gap, https://www.wipo.int/en/web/economics/innovation-gender-gap (last visited March 25, 2026).
89. Intersection of Generative Artificial Intelligence and Copyright: An Indian Perspective, ResearchGate (2026), https://www.researchgate.net/publication/382106745_Intersection_of_generative_artificial_intelligence_and_copyright_an_Indian_perspective.
90. Catherine D’Ignazio et al., Feminicide and Counterdata Production: Activist Efforts to Monitor and Challenge Gender-Related Violence, 3 Patterns (N Y) 100530 (2022).