Artificial intelligence and creativity

The present report builds on the previous report of the Special Rapporteur, on the digitalization of cultural heritage (A/HRC/58/60). It addresses the impact of artificial intelligence (AI) on cultural rights, in particular the freedom indispensable for creative activity, the right to artistic freedom and the right to enjoy the arts, as well as the right to the protection of the moral and material interests of authors. These are all important elements of the right to participate in cultural life, which States have undertaken to respect and protect in accordance with article 15 of the International Covenant on Economic, Social and Cultural Rights.

AI has expanded so rapidly and so extensively in the past decade that there has been no time to reflect on its implications. There is no doubt that AI tools offer amazing support for creative solutions that affect many aspects of society. AI tools can help to push the imagination and help people to visualize solutions for their well-being and enjoyment. They can amplify the exercise of the right to participate in cultural life.

Despite the definite genuine support that AI can offer for creative solutions, the bulk of AI creative tools today are hijacked by big technology corporations whose main interest is not the advancement of humanity, but rather economic profit. United Nations reports have repeatedly highlighted the need to look at how AI and AI tools are actually used currently in the hands of these private, profit-based powerful companies.¹

1. For an overview of the numerous reports on this topic, see A/HRC/56/45. ↩

There has certainly been a delay in taking measures to address the shortcomings of AI. Instead, the positive aspects of AI are pushed so much by marketing campaigns that the world’s collective eyes are not set on the shortcomings. High-tech corporations and investors assure us that AI is the next big development that will revolutionize our realities, support our dreams and improve our lives. Project after project, we are informed that machines will gradually perform all mundane tasks, leaving people with more space for the important ones, and that AI will make our decisions, even medical ones, and will ensure our comfort. The information is relentless, and the rapidly shifting nature of AI technologies prevents us from engaging with a clear head regarding the benefits and concerns these developments raise.² Without the time to reflect on these questions, the “inevitable” character of technological advancements is put forward, together with the urge to embrace, and not fight, AI.

2. Submission by Saudi Arabia. ↩

The High-level Advisory Body on Artificial Intelligence has noted: “When we look back in five years, the technology landscape could appear drastically different from today. However, if we stay the course and overcome hesitation and doubt, we can look back in five years at an AI governance landscape that is inclusive and empowering for individuals, communities and States everywhere. It is not technological change itself, but how humanity responds to it, that ultimately matters.”³ This is the basis for the focus of the present report on creativity.

3. Governing AI for Humanity, Final Report (United Nations publication, 2024), para. 216. ↩

The international human rights obligations of States prescribe the need to take specific and special measures to address and mitigate the shortcomings of AI. After consideration of the relevant work of the United Nations,⁴ including that of the Office of the High Commissioner for Human Rights⁵ and its B-Tech Project,⁶ the special procedures of the Human Rights Council,⁷ the International Labour Organization (ILO),⁸ the United Nations Educational, Scientific and Cultural Organization (UNESCO),⁹ the International Telecommunication Union,¹⁰ the Organisation for Economic Co-operation and Development,¹¹ the World Economic Forum,¹² the European Union,¹³ the Council of Europe¹⁴ and the African Union,¹⁵ and after consultation with civil society, academia and multi-stakeholder initiatives, the Special Rapporteur warns that the impacts of AI on creativity and cultural rights have largely been neglected. In preparing the present report, the Special Rapporteur also widely distributed a questionnaire in order to collect various views and experiences, to which 86 responses were received.¹⁶ The Special Rapporteur thanks all for their contributions.

4. See A/79/296; A/HRC/43/29; and United Nations, “Guidance of the Secretary-General: human rights due diligence for digital technology use”, May 2024. ↩
5. See A/HRC/39/29, A/HRC/44/24 and A/HRC/48/31. ↩
6. See www.ohchr.org/en/business-and-human-rights/b-tech-project. ↩
7. See, for example, A/73/348, A/74/493, A/76/151, A/78/310, A/HRC/32/45, A/HRC/38/48, A/HRC/41/41, A/HRC/46/37, A/HRC/47/52, A/HRC/49/52 and A/HRC/50/32. See also https://empresasyderechoshumanos.org/wp-content/uploads/2022/04/INFORMATION-NOTE-on-PP_LAC_EN.pdf. ↩
8. See www.ilo.org/artificial-intelligence. ↩
9. See www.unesco.org/en/articles/recommendation-ethics-artificial-intelligence; www.unesco.org/ethics-ai/en/eia; and https://unesdoc.unesco.org/ark:/48223/pf0000391566. ↩
10. See www.itu.int/en/action/ai/Pages/default.aspx. ↩
11. Organisation for Economic Co-operation and Development, OECD Guidelines for Multinational Enterprises on Responsible Business Conduct (OECD Publishing, Paris, 2023). ↩
12. www3.weforum.org/docs/WEF_AI_Procurement_in_a_Box_AI_Government_Procurement_Guidelines_2020.pdf. ↩
13. See https://digital-strategy.ec.europa.eu/en/policies/regulatory-framework-ai. ↩
14. See www.coe.int/en/web/artificial-intelligence. ↩
15. See https://au.int/en/documents/20240809/continental-artificial-intelligence-strategy. ↩
16. All submissions are available at www.ohchr.org/en/calls-for-input/2025/call-contributions-artificial-intelligence-and-creativity. ↩

The aim of the report is to ensure that public authorities have control over the future of AI, that they can regulate its production and use and that they are accountable for its impact, so that people can use it in a meaningful and self-empowering manner, without discrimination, fully retaining their right to express their creativity in all fields of human life. Workshops, discussions and plans must urgently be transformed into action; otherwise, humanity faces an impending avalanche of catastrophic consequences for cultural rights.

Autonomy and adaptability are the two elements that define and distinguish an AI system from other systems, including digitalized ones. According to the 2024 Council of Europe Framework Convention on Artificial Intelligence, Human Rights, Democracy and the Rule of Law, 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 may influence physical or virtual environments” (article 2). The 2024 Artificial Intelligence Act of the European Union articulates these two elements explicitly (article 3 (1)). The feature of autonomy is also implicit in the 2021 UNESCO Recommendation on the Ethics of Artificial Intelligence (article 2).

Several elements included in the definitions of AI suggest that it may potentially be used in creative processes. “Creativity is traditionally understood as the capacity to produce works that are both original and meaningful.”¹⁷ Creativity is part of everyday life, and combines the use of imagination, past experiences and logic to find new ideas or create new elements. It is not confined to the arts, but spans science, engineering, literature, business innovation and everyday problem-solving. “What sets human creativity apart from machine-generated output is its subjective depth, contextual grounding, and capacity for narrative, empathy, and ethical judgement. Philosophically, creativity is also tied to consciousness and agency.”¹⁸ Creativity is more than productivity: it includes emotional resonance, perception, talent, narrative and the ability to break patterns and engage with uncertainty, skills that can be applied in all spheres of life and all types of activities. “Human creators make decisions not only based on data, but also on aesthetics, intuition, memory and emotion. These factors are deeply embedded in human experience and cannot be replicated by machines, no matter how advanced the algorithms.”¹⁹ AI is a creative tool; it can be used creatively, but is not a source of creative soul.²⁰

17. Submission by Saudi Arabia. ↩
18. Ibid. ↩
19. Ibid. See also submission by Guillaume Dumas, Antoine Bellemare, Suzanne Kite and Karim Jerbi. ↩
20. Submission by Christian Steinau, Julian Stalter, and Lucas Hagin, pp. 3 and 4, referring to Sean Dorrance Kelly, “What computers can’t create: why creativity is, and always will be, a human endeavor”, MIT Technology Review, vol. 122, No. 2 (2019). ↩

It is vital to distinguish explicitly and clearly content that is human-driven from that which is purely synthetic.²¹ In addition, it is essential to recognize that purely synthetic material is not creative, but rather an automatic addition, even when the quality is very good. AI has also been referred to as “average Internet”, reflecting the fact that it is not creative as such, but can learn only from what existed before and infer what will come next based on the most likely (or the most mainstream) occurrences. Creativity is an indispensable part of the human experience and of individual and shared identities and cultural resources. It must be firmly guarded. This can happen only by adopting a human rights, specifically, a cultural rights, approach.

21. Submission by Saudi Arabia. ↩

Using AI tools for creativity-related functions challenges the integrity and authenticity of human creative expression, thereby putting at risk several dimensions of cultural rights, not least the right to freely participate in cultural life.

Human expression is not understood only as outcomes (a book, a renovated house, a scientific discovery) but is celebrated as a process of the human mind that includes conscious decisions and emotional engagement. Writing a poem, composing a musical piece or cooking a dish are not only about the outcome, but about the processes that involve part of one’s past, one’s own ideas and the preferences of one’s family. Such processes say something specific about human beings to the world in a way that an AI-generated book or musical tune will never do. This calls for reflection on the need to value and preserve human creativity, which now, more than ever, is challenged by AI tools.²²

22. National Human Rights Council of Côte d’Ivoire. ↩

The use of AI tools also blurs the distinction between works produced by humans and those wholly or predominantly generated by AI. In the publishing sector, for instance, major platforms such as Amazon present AI-generated books as works by authors, even though they are merely counterfeit or low-quality texts that may even appropriate the identities of real authors.²³ Transparency about AI generation is essential to ensure the cultural rights of both the creators and the receivers. However, transparency, even when included in legislation, is not monitored.²⁴

23. Submission by Association nationale des éditeurs de livres, p. 2, referring to Agence France Presse, “Les livres écrits avec ChatGPT envahissent Amazon, 10 May 2024. ↩
24. See Partenariat Interprofessionnel du Livre et de l’Édition numérique, “Charte pour une utilisation responsable de l’intelligence artificielle dans le secteur des écritures et du livre”, available at https://pilen.be/charte. See also submission by the Alan Turing Institute, p. 7. ↩

Increasing the dependence of individuals on AI systems also has long-term effects on human creativity and, consequently, on cultural rights. In some sectors, it seems difficult to fight this dependence. Researchers at Bridging Responsible AI Divides noted: “creators and businesses are under an increasing pressure of using AI at all costs, driven by hype and competitive anxiety. The risk associated with this is the reduction of creative processes to productivity and efficiency metrics.”²⁵ Such an acceleration in the pace of production inevitably limits exploration of and deep reflective practices for genuine cultural innovation.²⁶

25. Submissions by Bridging Responsible AI Divides researchers, p. 3, and Caterina Moruzzi, p. 2. ↩
26. Submission by Jorge Caballero Ramos, p. 1. ↩

The use of AI tools amounts to a “delegation of thought” for generative production accompanied by “broad de-responsibilization and intellectual laziness stemming from ‘digital addiction’”.²⁷ This dependence results in a loss of knowledge, skills and self-confidence, which is a major issue.²⁸ Consequently, the transmission of creative knowledge and skills is interrupted, and the ability of individuals to fully participate in cultural life as creators is inevitably reduced. AI systems capable of generating new content, ideas or data that mimic human creativity risk damaging or destroying the transmission of skills to newer generations.²⁹ How can a PhD researcher learn to engage in high-quality research, if not by repetitive searches and critical assessment of the nuances? Creation happens not only through success, but also through failure. How can artisans connect to their tools, if not by regularly using, cleaning and taking care of them? The “mindless, repetitive actions” that AI purportedly will take care of to allow us to be more creative are often part of the processes of creativity. Such actions bring joy and pain; they are all part of the human life cycles that generate creation and experience.

27. Submission by Michele Pasquale, p. 4. ↩
28. Submission by Jeanette Folk, p. 3. See also, the Massachusetts Institute of Technology research on ChatGPT and cognitive decline, available at www.media.mit.edu/publications/your-brain-on-chatgpt/. ↩
29. Submission by Harry H. Jiang, p. 2. ↩

Generative AI tools also raise specific concerns regarding children’s creativity. Ethical and human rights risks include the loss of autonomy, the datafication of play, normative bias and reduced opportunities for challenges.³⁰ The AI-driven environment may also limit spontaneous play and imagination, key drivers of child development and self-expression. Such AI tools must be subject to repeated evaluation.

30. Submission by Nomisha C. Kurian, pp. 3 and 4. ↩

The collection of data by platforms poses serious concerns about a new form of colonialism that extracts not natural resources, but rather data resources. Big companies grab data from every source, open or not, usually without any consultation or consent, and without communities being informed about all the consequences.³¹ Data inventories for AI systems, including cultural data, are rarely built in participatory, transparent ways, nor are they guided by human rights frameworks. The amount of energy used to gather and sustain such data is catastrophic for the environment. Moreover, most of those platforms are governed by a small number of actors, often from the Global North, whose worldviews and economic interests shape the systems’ design, reinforcing existing biases and structural inequalities. Never before have communities envisaged that their lives would become data to be used in such an unlimited and opaque way.

31. See UNESCO, “Operational guidelines on the implementation of the Convention in the Digital Environment”, 2017. In para. 8.9, it is stated that respect for the right to privacy is a sine qua non condition for the creation, distribution and accessibility of diverse cultural expressions. ↩

Such data is kept by big corporations and used in any way they see fit, with profit being the main goal. The visions, values and cultural references embedded in that data are not interpreted by the source communities, and the ethics of their usage are not discussed; creative works are used without any consent or acknowledgement, which poses serious questions regarding the respect due to cultural rights. Music is taken in a piecemeal way without any regard for the original owner or its meaning or importance. Communities’ designs are taken without any consent and are sometimes altered in an arbitrary way. Drawings, pictures, photographs and any materials – whether sacred and important, or not – are stored and considered meaningful only as parts of this profit-making endeavour. For example, ChatGPT has enabled users to transform Internet memes or photographs into the distinct style of the founder of Studio Ghibli, Hayao Miyazaki, even though he has been explicitly opposing AI for decades.³² Individuals’ records, stories and vulnerabilities are stored without consent or emotional engagement, only to be disseminated when and as the new “owners” decide, not for the benefit of humanity, but for economic benefits that, it is hoped, may also benefit humanity. There is no effective right to appeal and no right to ask that data be erased, not reproduced or not stored in the first place.

32. See www.independent.co.uk/arts-entertainment/films/news/hayao-miyazaki-studio-ghibli-ai-trend-b2723358.html. ↩

International human rights law has not been effective in restricting AI systems from indiscriminately collecting data, even despite new legal instruments at the international level, including the Council of Europe Framework Convention on Artificial Intelligence and Human Rights, Democracy and the Rule of Law, which establishes legally binding obligations with respect to compliance of AI activities with human rights; the Artificial Intelligence Act of the European Union, which prohibits a number of AI-related practices and establishes transparency requirements; and the Continental Artificial Intelligence Strategy of the African Union, which highlights the risk of appropriating and misrepresenting Indigenous knowledge. The Special Rapporteur is concerned that the loss of cultural rights in collecting data is not the focus of scrutiny by legislators and human rights organizations, even though data collection is advancing at a rapid pace.

Several court proceedings are currently under way, but they involve big companies fighting over who will control and benefit from the data of the creative work of individuals. In the United States, the concept of “fair use” is invoked by AI developers to justify training their models on billions of works (texts, images, music) without compensating rights holders. Fair use is an affirmative defence in copyright law that permits limited use of copyrighted material without permission from the rights holder, typically for purposes such as criticism, commentary, news reporting, teaching, scholarship or research. In February 2025, a court in the United States of America rejected an AI developer’s fair use defence, arguing that the use was commercial in nature and not transformative, as it did not substantially alter the original content to create new meaning or expression.³³

33. United State of America, District Court of Delaware, Thomson Reuters Enterprise Centre GMBH v. Ross Intelligence Inc., Case No. No. 1:20-cv-613-SB, Memorandum Opinion, 11 February 2025. ↩

Similar cases have emerged in the music industry. In the United States, Sony Music, Universal Music Group and Warner Music Group are suing the start-ups Udio and Suno, alleging that they used copyrighted works to develop their AI models, which generate music based on user prompts. In India, a group of major newspaper and book publishers has brought a case before the Delhi High Court, claiming that OpenAI is using their content without authorization to train ChatGPT.³⁴ In the United Kingdom of Great Britain and Northern Ireland, Getty Images has filed a lawsuit against Stability AI, alleging that the developer used millions of its copyrighted images, along with metadata, without permission, to train its image-generating AI model, Stable Diffusion.³⁵

34. See Delhi High Court, Ani Media Pvt. Ltd vs Open Ai Opco Llc, Order, 28 March 2025 (case pending). ↩
35. See High Court of Justice, Business and Property Courts of England and Wales, Intellectual Property, Getty Images and others v. Stability AI, Judgement, 14 January 2025. ↩

European copyright law, unlike United States copyright law, does not recognize the concept of fair use, but text and data mining are permitted on an exceptional basis. The European Union Directive on copyright in the Digital Single Market, revised in 2019, allows “reproductions and extractions made by research organisations and cultural heritage institutions in order to carry out, for the purposes of scientific research, text and data mining of works or other subject matter to which they have lawful access” (article 3). Article 4 permits other actors to make “reproductions and extractions of lawfully accessible works and other subject matter for the purposes of text and data mining.” Rights holders may opt out, allowing them to prohibit text and data mining.³⁶ The implementation of such opt-outs are very problematic, however, as the companies remain very opaque and the lack of familiarity with data content on the part of individuals and communities is widely exploited.³⁷ Clearly, AI companies do not want to enter into licensing agreements, nor are they transparent about the copyright-protected works used for training. “This has led to a total lack of communication between AI companies and rights holders, where rights holders’ requests to negotiate and discuss are ignored.”³⁸

36. Alexandra Bensamoun, “IA et culture: de la création aux données”, in Intelligence artificielle, culture et médias, Véronique Guèvremont and Colette Brin, eds., (Canada, Presses de l’Universite Laval, 2024), p. 337. ↩
37. Submission by European Coalitions for Cultural Diversity, p. 1. ↩
38. Submission by Society of Audiovisual Authors, pp. 2 and 3. ↩

Copyright regimes are especially ill-suited to protecting traditional cultural expressions, and AI tools could exacerbate their vulnerability. The mandate holder has repeatedly discussed the need to make copyright provisions effective.³⁹ In India, for example, threats to traditional cultural expressions already come from decreasing demand for them, the influence of middlemen and intermediaries operating copyright regimes, and many other factors. “This is exacerbated by threats from AI-generated art, which imitates [artisans’] style and processes without recognition or compensation. This form of traditional knowledge and art receives very superficial protection under the Indian Copyright Act due to the communal nature and enforcement challenges.”⁴⁰

39. See A/HRC/28/57 and A/HRC/58/60. ↩
40. Submission by Prakhar Singh, Samrudh Chirkankshit Bulani, Hargun Kaur, Amishi Jain and Shatrupa Sharma of the Center for Advanced Studies in Cyber Law and Artificial Intelligence, p. 7. ↩

It is a matter of urgency that clear and effective limitations be created for companies with respect to data sourcing. In the absence of effective protection, creatives have embedded in their works invisible modifications known as “filters” or “algorithmic camouflage” to prevent AI systems from learning from them.⁴¹ But these measures cannot be the way forward. Users of AI systems must be able to understand how they are developed, including the sources of data, the methods of training, the actors involved and the intended uses. Transparency across the entire AI life cycle and standardized protocols to effectively implement an opt-in model on the Internet must be put in place by States. A collective compensation fund, financed by a tax on the revenues of large AI platforms, or harmonized standards in accordance with legislation, as developed in the European Union by the European Committee for Standardization and the European Committee for Electrotechnical Standardization, would also be solutions.⁴² There is an urgent need to act to protect cultural rights from uncontrolled data grabbing. There is no time to continue with mere reflection on these issues, nor time to plan lengthy and complex reforms.⁴³

41. Submission by Benjamin Harbakk, p. 6. ↩
42. Submissions by Pau Aleikum Garcia of Domestic Data Streamers, p. 3; the Alan Turing Institute, p. 4; and the Writers’ Guild of Great Britain, p. 2. ↩
43. Submission by Bureau Malien du Droit d’Auteur, p. 1. ↩

In promoting only the creative forms that will sell, AI restricts cultural diversity and promotes homogenization. AI models remain fundamentally designed to reproduce statistically average patterns from their training data, reinforcing the most common styles. “AI systems trained on dominant cultural inputs tend to reproduce dominant aesthetics. The effect is subtle but significant: over time, global platforms begin to favour homogenized output, marginalizing regional, Indigenous, or non-Western modes of expression.”⁴⁴ This leads to the standardization of creativity. In addition, the average patterns easily promote pieces that defy the search for quality in art. The quality gradually and continuously falls as successive generations of AI are trained on the previously average content, intensifying the standardization of styles and ideas and further reducing originality and diversity. The same can be said of redirection by AI towards conformity with the mainstream, which reduces the prevalence of non-mainstream opinions.⁴⁵ In parallel, reducing the range of diversity to which people are exposed also weakens their ability to compare and appreciate a range in quality; as a consequence, their freedom to know and to choose is also significantly reduced.

44. Submission by Davar Ardalan and Amir Banifatemi, p. 2. ↩
45. Daragh Murray and others, “The chilling effects of surveillance and human rights: insights from qualitative research in Uganda and Zimbabwe”, Journal of Human Rights Practice, vol. 16, No. 1, 31 July 2023. ↩

The extremely rapid development of AI systems that generate text, images, videos and music accelerates this downward spiral of quality in creativity. AI tools adapted to specific forms of human creativity struggle to reproduce the nuances and variations that artists introduce into works. In Indian classical music, the microtonal variations known as shruti and raga cannot be reproduced by AI tools such as GaMaDHaNi and NaadSadhana, thus failing to capture the emotional essence and depth of what artists are trying to convey through the art form. The widely popular dance form from Tamil Nadu has its own set of codified gestures and rhythmic sequences that become a struggle for AI models such as Natya.AI. AI robs the dance form of its emotional expression. AI technologies cannot replicate the cultural significance and emotions.⁴⁶

46. Submission by Prakhar Singh, Samrudh Chirkankshit Bulani, Hargun Kaur, Amishi Jain, Shatrupa Sharma and the Center for Advanced Studies in Cyber Law and Artificial Intelligence, pp. 6 and 7. ↩

The opacity surrounding the criteria or data used by big technology companies in order to suggest certain works to users makes the future bleaker. This lack of transparency is a common feature of most major cultural content distribution platforms. A ray of light is TV5MondePlus, which uses AI to promote and support works that might otherwise be overlooked. By adapting the codes and interface logic of mainstream platforms, it actively leverages AI to promote cultural diversity and ensure greater visibility for a wider range of voices.⁴⁷

47. See https://tv5mondeplus.com. ↩

The increased use of AI also affects languages negatively. Cultural works can now be subtitled, dubbed or translated by AI systems alone – a feature referred to as automatic or automated translation – rather than through human intervention. AI tools can be used for language preservation and revitalization: for example, native speakers can be recorded so that audio content in their language can be generated, thereby facilitating transmission to new generations. However, such tools currently seem unable to faithfully reflect what creators express, including intention, sensitivity and nuances, as well as irony and sarcasm. “Cultural understanding and nonverbal communication are crucial in interpreting. Technology cannot adequately convey cultural nuances, body language, idiomatic expressions, and context-dependent meanings, leading to possible misinterpretations and subsequent harm and infringement of rights.”⁴⁸ Automated translation indeed uses statistics and probabilities to suggest text in the target language; there is no real analysis or understanding of the conveyed message. Machines lack human judgment and – unlike the professional licensed interpreters acting under applicable laws – cannot be held legally accountable.⁴⁹ While AI facilitates access to documents, in a sense democratizing access to information, it also leads to linguistic impoverishment, that is, a reduction of actively used vocabulary, a simplification of syntactic structures, stylistic homogenization and the gradual erasure of regional or social language variations. Efforts to improve AI tools in languages other than English are very slow, as they are not as profitable. Minority and Indigenous languages are especially affected.

48. Submission by the International Federation of Translators, p. 3. ↩
49. Submission by the Czech Association of Interpreters and Translators, p. 3. ↩

All these issues severely impact cultural rights, particularly the right to participate in cultural life in one’s own language, as well as the rights to create and distribute works in any given language. This creates fears of exclusion and the disappearance of linguistic diversity from the digital environment and, in the long run, in the physical environment too.

Prevailing inequalities limit the ability of certain individuals and groups, particularly those belonging to marginalized communities, including those living in poverty and in remote areas, to use AI tools in a meaningful and self-empowering manner to create and produce their works in all fields of human creativity. For many creators, AI tools do not support the dissemination of their works or make them more accessible to the public.

While the digital divide persists in terms of material access to digital devices and connectivity, access to the knowledge and skills required to use them, and the benefit derived from digital technologies,⁵⁰ a creative divide is worsening existing inequalities.⁵¹ In addition, the protections offered by existing or emerging legal frameworks are inconsistent. Brazil, in its submission, noted that “creators from low-income backgrounds and the Global South suffer disproportionately from the negative impacts of AI disruption on cultural economies, while lacking access to the legal protections enjoyed in more privileged contexts.”⁵²

50. Submission by Fundación para la Democracia, p. 1. ↩
51. See the Fair Culture Charter, principle 6, which states that “Equitable access to digital tools, digital literacy, skills, and capacities, along with allocating resources to bridge digital gaps, are critically needed as well”. ↩
52. Submission by Brazil, p. 2. ↩

Ultimately, the creative divide amplifies the existing imbalance in the representation of diverse cultures in AI-generated content: “Those with access to compute power, data infrastructure, and dominant languages disproportionately shape the outputs and aesthetics of generative AI, often marginalising other cultural perspectives”. So far, the United Nations bodies have been very slow to address AI attacks on creativity. A clear emphasis on such violations is important, as “without careful attention to these asymmetries, AI risks amplifying existing inequities in whose creativity is recognised, valued, and preserved.”⁵³

53. Submission by Eva Nieto McAvoy, p. 3. ↩

The recommendation by UNESCO on the ethics of AI calls on AI actors to “make all reasonable efforts to minimize and avoid reinforcing or perpetuating discriminatory or biased applications and outcomes throughout the life cycle of the AI system to ensure fairness of such systems.”⁵⁴ However, there is no concrete suggestion regarding measures, nor is there any monitoring mechanism.

54. UNESCO, “Recommendation on the ethics of artificial intelligence”, 2022, para. 29. ↩

AI tools are not neutral; they are the products of political, technical, linguistic and economic decisions.⁵⁵ AI cultural outputs reflect dominant norms and their recommendation systems produce distorted results, either because they are shaped by profit-driven models or because they reflect data gaps. As AI tools reproduce the data that they have been fed in an uncritical and unchallenged manner, it is no surprise that their content is partial, stereotypical and discriminatory. They generate content at a scale and pace not seen before and in a manner that is often unrepresentative of the diversity of cultural identities, heritages or languages. This cultural bias is part of a broader ethical challenge: AI systems – whether generative, predictive or decision-making – tend to reinforce and even exacerbate existing social, cultural, economic and political inequalities.

55. Submission by Nicolás Madoery, FUTURX, p. 2. ↩

Particular attention must be paid to the effect that AI has on minorities, Indigenous Peoples and other marginalized groups. The underrepresentation of data from these groups in the training of models results in biased outputs that fail to accurately reflect the identities of these groups. Their cultures, values, knowledge, narratives, aesthetics and diverse artistic expressions are either absent⁵⁶ or misrepresented in AI-assisted or AI-generated creations. In sectors or platforms concentrating large volumes of AI-produced works, this leads to the invisibility of works representing these cultures, perpetuating pre-existing inequalities. Underrepresentation can also lead to stereotypical representations, offering folklorized or stigmatizing views of these groups, reinforcing prejudices, disrespecting their cultural identities and harming their dignity.⁵⁷ All these negative elements have been gradually exposed in the literature on AI in the past few years, to such a degree that scholars wonder whether AI is becoming the new colonizer of Indigenous Peoples.⁵⁸

56. Submission by Anna Su, p. 2. ↩
57. Submission of the Human Rights Ombudsman of Guatemala. ↩
58. Jason Edward Lewis, ed., Indigenous Protocol and Artificial Intelligence, position paper (Honolulu, 2020). ↩

Decontextualization is evident in the fashion industry. The use by AI of Indigenous garments and motifs is reduced to mere aesthetics, which strips those garments and motifs of the cultural memory that grounds their meaning. Prints “inspired” by Maori or Nigerian heritage motifs, taken without any acknowledgment of where those motifs came from or of who designed them, and without consultation with the source community, violate cultural rights and potentially propagate harm. AI systems have disassociated designs from their social and historical lineages, converting them into commodified fragments in a broader process of algorithmic consumption.⁵⁹ AI tools could be used with the active participation and consent of the communities, which would ensure the coherent use of the designs in line with their real meaning and significance.

59. Submission by Indira Boutier, p. 7. ↩

Initiatives to increase the availability of data related to marginalized groups in digital environments without the consent of the source community carry risks. They can result in a loss of control over narratives and cultural representations,⁶⁰ as well as cultural appropriation, whereby “ancestral knowledge, sacred art forms and traditional expressions become training data.”⁶¹

60. Submission by Centro de Investigación y Docencia Económicas and Artículo 19 Oficina para México y Centroamérica, p. 3. ↩
61. Submission by Brazil, p. 2. ↩

In some emerging practices, the groups concerned are involved at every stage of projects that affect them. For example, the Creative Labour and Critical Futures research cluster works with minorities.⁶² The Mila – Quebec Artificial Intelligence Institute, led by Michael Running Wolf, uses AI to document and revive Indigenous languages in cooperation with local communities to the extent that they agree their data may be used.

62. Submission by Creative Labour and Critical Futures, p. 5. ↩

AI-generated content may also reinforce stereotypical representations of women. These biases stem from training data sets that underrepresent women’s voices, experiences and contributions, or that overrepresent them in traditional or objectifying roles. “The underrepresentation of women in AI development and leadership roles can further lead to the creation of socio-technical systems which fail to consider the diverse needs and perspectives of all genders, once again perpetuating stereotypes and gender disparities.”⁶³ As a result, AI systems can contribute to the rendering invisible of the diversity of women’s identities and roles in societies. This impedes the exercise of women’s cultural rights.

63. UNESCO and International Research Centre on Artificial Intelligence, “Challenging systematic prejudices: an investigation into gender bias in large language models” (Paris, 2024), p. 5. ↩

Generative AI tools pose specific challenges for women and girls, notably by reinforcing gender-based discrimination and enabling new forms of harm to their dignity and integrity. For instance, “text-to-image models can easily generate images of women in situations they did not consent to being in, thus creating a more realistic vector of image-based abuse; creating synthetic histories is a new vector of technology facilitated gender-based violence harm.”⁶⁴ One of the most serious threats is the use of AI to produce non-consensual deepfake pornography, which can be used to harass, intimidate or silence women. These technologies facilitate new forms of online abuse, exacerbating pre-existing patterns of violence against women in digital environments.

64. UNESCO, “Your opinion doesn’t matter, anyway: exposing technology-facilitated gender-based violence in an era of generative AI” (Paris, 2023), p. 7. ↩

The impact of AI on the creative sector labour market, particularly the replacement of lower-skilled jobs by AI tools, may also have a specific effect on women. Women are often overrepresented in precarious forms of employment, which are those most likely to disappear with the growing use of AI in this sector. This trend increases economic insecurity for women and may lead to their exclusion from creative fields or, at the very least, a reduction in their representation. Moreover, women generally have more limited access to technology than men, are often less trained in its use and remain underrepresented in AI-related jobs.⁶⁵ As a result, they are more affected by the creative divide caused by AI.

65. This is partly due to the working environment in the technology sector, which is often marked by gender bias, a lack of inclusivity and barriers to advancement for women. See: UNESCO, “Challenging systematic prejudices”, p. 20. ↩

Creativity is also affected by the essential role of AI in the distribution and dissemination of and access to information in the digital environment. AI is extremely effective in the analysis of user behaviour and in recommending works that match individuals’ preferences, enabling individuals to be manipulated into making specific choices. For example, Netflix suggests what to watch next; Google suggests what to buy; Meta selects specific news items; AI facilitates access to certain works through automated translation or subtitling in certain languages. On music platforms, a user’s behaviour is tracked and data compiled on how often a track is played, whether it is added to one or more playlists and whether the user skips to the next track before it finishes. AI systems use this “big data” – that is, continuously produced, large-scale digital data – to create user profiles and analyse users’ behaviour (preferences, habits, cultural interests, etc.) without their meaningful consent. Through AI, platforms analyse this data and apply filters. They compare the behaviour of users who have expressed similar tastes (collaborative filtering) and the characteristics of previously liked content (content-based filtering) in order to suggest similar works. This can be positive, as it allows people to discover their preferred cultural materials, or it can be negative, resulting in the creation of a bubble of homogeneity and bias.

The Billie Eilish song “Ocean Eyes”, which went viral on SoundCloud in 2015, or the Lil Nas X song “Old Town Road”, which went viral on TikTok in 2018, are well-known examples of artists gaining exposure thanks to algorithms and recommendations. AI systems can also facilitate the discovery of more niche works. AI can therefore help to connect certain creators with certain audiences, which has a substantial impact on the right to enjoy the arts.

However, information about creative tools and elements that are not highlighted by algorithms is less accessible, and although not all algorithms are based on AI, this is increasingly the case. The use of AI for algorithmic development accentuates the shortcomings of algorithms for creativity. The algorithmic development selectively allows for some cultural creators to be exposed to the public and some not. Audiences are potentially presented with a reduced diversity of works. All such AI-generated choices and the related obstacles are, in essence, not made by the individuals themselves, but are a reflection of strategic economic interests to maximize user engagement and establish market dominance. For example, a streaming platform also involved in content production (such as Netflix) will tend to highlight the works that it has produced. This is referred to as algorithmic self-promotion or algorithmic favouritism. Platforms also steer audience attention towards cultural content the consumption of which maximizes their profit.

Recommendation algorithms made more effective by AI tools have a direct effect on the diversity of creative works in the digital environment, which is regarded as a public space.⁶⁶ Creative works that do not conform to dominant or popular styles may be disadvantaged or even completely ignored. Creative expressions from non-dominant cultures may also be entirely excluded by recommendation algorithms.⁶⁷ Works by independent artists, local artists or groups addressing topics of interest to smaller groups may be rendered invisible. This leads to clear discrimination when certain types of works or works by certain creators are systematically disadvantaged. AI systems can also be used to optimize recommendation algorithms as tools of censorship, which is at odds with the right to freedom of artistic expression and creativity. “Invisibilization” may affect works that address sensitive or polarizing subjects or artists that represent certain cultures or schools of thought. Don’t Delete Art is an example of a group that draws attention to the need to restore public access to artistic expressions that have been excluded from public discourse. It republishes on its own website artistic content that has been removed by platforms and provides artists with a list of hashtags to avoid in order to prevent censorship.⁶⁸ However, measures must also be taken by States, not only by civil society.

66. Submission by Laurence Cuny, p. 2. ↩
67. Submission by European Coalitions for Cultural Diversity, p. 1. ↩
68. Laurence Cuny, “Être artiste aujourd’hui: quelques défis de protection de la liberté artistique dans l’espace numérique”, Lex Electronica, vol. 28, No. 4 (2023), p. 38. ↩

The fact that a small number of platforms disseminating cultural content dominate the global market is concerning. OpenAI, ChatGPT, Google Gemini, Meta AI and Microsoft Copilot control the infrastructure and the choice of cultural content they share. They decide which voices are indexed and which are buried, generating an information asymmetry⁶⁹ that may be affecting the discoverability of works. In essence, by allowing the uncritical use of AI, control over human realities has been surrendered to machines, or, rather, to the big technology companies that control such machines. International human rights law is clear: States must protect human rights by taking action to limit obstacles to their full realization.

69. Submission by Pau Aleikum Garcia, p. 1. ↩

The interest of large corporations in the processes and outcomes of creativity has had a deep effect on cultural workers.⁷⁰ New professions are being created that combine artistic and digital skills, such as algorithmic curators or content editors.

70. See Janine Berg and others, “Generative AI and the media and culture industry”, ILO research brief (2025). ↩

However, the forms of creativity that do not use AI systems suffer. Artists’ renumeration is at stake, notably because platforms themselves invest in inexpensive AI-generated creation and promote their own content to reduce copyright payments.⁷¹ A recent study highlighted the “potential cannibalisation of creators’ revenue streams due to the substitution of human works by [generative] AI outputs”.⁷² Within only a few years, creators could lose nearly a quarter of their income: “under current conditions, this market penetration by [generative] AI outputs could put 24% of music creators’ revenues at risk by 2028”.⁷³

71. See Feriel Mestiri, “Des morceaux de musique créés par l’IA inondent les plateformes de streaming”, Radio Télévision Suisse, 1 April 2025. ↩
72. International Confederation of Societies of Authors and Composers, Study on the Economic Impact of Generative AI in the Music and Audiovisual Industries, November 2024, p. 41. ↩
73. Ibid., p. 71. ↩

Many sources report that the destruction of the professional fabric is evident.⁷⁴ In this respect, “AI presents a double-edged sword. While it promises productivity gains, efficiency, and economic growth, it also raises concerns about job displacement, skill gaps, surveillance, and the erosion of workers’ rights”.⁷⁵ Reports suggest that people working in creative sectors (illustration, photography, writing, editing and translation, among others) are seeing reduced demand for their services and, when their services are contracted, they are often relegated to merely reviewing what the machine has produced.⁷⁶ The competition faced from AI tools is leading to fears for the disappearance of certain professions, increased precarity in cultural jobs, lower pay for artists and other cultural professionals, and a devaluation of human creativity.⁷⁷ These developments coincide with questions as to the quality of the work. In a recent survey of creative workers in the United Kingdom, 73 per cent believe that AI is changing the quality of work in the creative industries.⁷⁸

74. See Octavio Kulesz, “Dialogues interdisciplinaires: repenser la culture à l’ère de l’IA”, International Observatory on the Societal Impacts of AI and Digital Technology and IVADO, May 2025, p. 16. ↩
75. Submission by Khalid Mahmood, Labour Education Foundation, Pakistan. ↩
76. Submission by Julija Kalpokiene. ↩
77. Submissions by Observatorio del Impacto Social y Ético de la Inteligencia Artificial, p. 1; Christian Steinau and others, p. 3; and the Alan Turing Institute, p. 2. ↩
78. Submission by Equity, pp. 2 and 3. ↩

Creators are aware that their works, when accessible in the digital environment, can be used by AI to generate new content, and that inexpensive copies or imitations are presented to the public in lieu of their work. Without attention to creativity or quality, the incentives for human creators to create more works are significantly reduced.⁷⁹ If regulatory frameworks are not updated to match the development of these technologies, there is a further risk that many creatives will cease to share their work for fear of it being used without their consent.⁸⁰

79. Submission by Society of Audiovisual Authors, p. 1. ↩
80. Submission by the Writers’ Guild of Great Britain, p. 1. ↩

Some artists are asked to sign contracts for the use of their voice or image on unfavourable terms, which include the purchase of their data on a buyout basis. “The one-off payments offered to performers who engage in generative AI work often do not reflect the fact that their image, voice or likeness may be used forever and on thousands of different projects”,⁸¹ or that this unlimited use might not be anchored in informed consent, control and fair compensation.

81. Submission by Equity, p. 3, as well as p. 9, on diverse regimes of personality or image rights applying to performers. See also submission by the International Federation of Actors, p. 5, on negotiation efforts to better protect the use of their members’ visual likenesses and voices for the creation of digital replicas. ↩

These transformations have, or will have, in the short and medium term, impacts on the economic and social rights of artists, as well as their socioeconomic status. Their artistic freedom is therefore under threat, as is their right to take part in cultural 71 See Feriel Mestiri, “Des morceaux de musique créés par l’IA inondent les plateformes de streaming”, Radio Télévision Suisse, 1 April 2025. 72 International Confederation of Societies of Authors and Composers, Study on the Economic Impact of Generative AI in the Music and Audiovisual Industries, November 2024, p. 41. 73 Ibid., p. 71. 74 See Octavio Kulesz, “Dialogues interdisciplinaires: repenser la culture à l’ère de l’IA”, International Observatory on the Societal Impacts of AI and Digital Technology and IVADO, May 2025, p. 16. 75 Submission by Khalid Mahmood, Labour Education Foundation, Pakistan. 76 Submission by Julija Kalpokiene. 77 Submissions by Observatorio del Impacto Social y Ético de la Inteligencia Artificial, p. 1; Christian Steinau and others, p. 3; and the Alan Turing Institute, p. 2. 78 Submission by Equity, pp. 2 and 3. 79 Submission by Society of Audiovisual Authors, p. 1. 80 Submission by the Writers’ Guild of Great Britain, p. 1. 81 Submission by Equity, p. 3, as well as p. 9, on diverse regimes of personality or image rights applying to performers. See also submission by the International Federation of Actors, p. 5, on negotiation efforts to better protect the use of their members’ visual likenesses and voices for the creation of digital replicas. life, to create and to make their work accessible to the public. AI may reduce the ability of artists to access the public sphere. III. Placing cultural rights and creativity at the heart of artificial intelligence policy

Measures must be taken as soon as possible to prevent the further loss of data, the flooding of platforms with inexpensive substitutes and copies of creative outputs and the erosion of cultural rights by the uncritical use of AI. Currently, international law does not have the teeth to address such challenges. The norms that provide the elements of an international legal framework applicable to AI and human creativity originate mainly from instruments focused on three areas: artificial intelligence, cultural diversity and intellectual property. The Special Rapporteur referred to these legal systems in her report on digitalization (A/HRC/58/60). Most of these instruments are legally non-binding and share a common feature in that, while they refer to fundamental rights and freedoms as a basis for setting objectives, principles or commitments, they fail to include explicit references to cultural rights. For example, although the UNESCO recommendation on the ethics of AI is an instrument strongly rooted in a human rights-based approach, it is non-binding, and both the 2024 revised AI principles of the Organisation for Economic Co-operation and Development and the 2024 Global Digital Compact also refer to rights, but not cultural rights, and are also not binding.

Several international political declarations refer to the human rights implications of AI, but have not been followed by concrete action. Real political will is important in addressing the violation of cultural rights by AI. States need to set limits and respect their human rights obligations when it comes to technological progress. Transparency and copyright are either not included in recent legislative initiatives, such as the Council of Europe Framework Convention on Artificial Intelligence and Human Rights, Democracy and the Rule of Law or, when they are, as in the Artificial Intelligence Act of the European Union, implementation guarantees are missing. Concerns have been raised that the draft code of practice for general purpose AI models, which was designed to accompany the Artificial Intelligence Act and is aimed at clarifying obligations with respect to transparency and copyright, will possibly weaken the text of the Act.⁸² Implementation of the legislative frameworks is important.

82. Submissions by the European Coalitions for Cultural Diversity, p. 2; the International Federation of Actors, p. 3; and the Alan Turing Institute, p. 5. ↩

Multinational enterprises are undoubtedly the main winners in the AI revolution.⁸³ The Guiding Principles on Business and Human Rights contain a recognition that all business enterprises have “an independent responsibility” to respect human rights and that, in order to do so, they are required to exercise human rights due diligence to identify, prevent, mitigate and account for how they address impacts on human rights.⁸⁴ The requirement of due diligence and the need for continuous evaluation of the impact of AI on human rights, including through impact assessments, have been repeatedly stressed, not least in paragraph 25 (a) of the Global Digital Compact.

83. See Véronique Guèvremont and Maxime Mariage, Fair Culture – A Key to Sustainable Development, German Commission for UNESCO (2021), p. 30. ↩
84. See A/HRC/59/53, para. 44. ↩

The legal obligation to guarantee respect for human rights, including by private persons and companies, still falls on the shoulders of States, and they must take urgent immediate measures to preserve the creativity of all and limit the obstacles that AI poses to the various elements of this right (see A/HRC/59/32). Using the cultural rights approach is essential.

Adopting a cultural rights approach is essential to addressing the shortcomings of AI. So far, all initiatives on AI, including legal ones, have, unfortunately, ignored cultural rights.

Cultural rights relate to human creativity in all its diversity and to the conditions for it to be exercised, developed and made accessible. They protect the development and expression of cultural identities. At their core, they empower individuals and groups to express their humanity, their worldviews and the meanings they attribute to their existence and development through various forms of human expression. Such forms of expression can include the arts, cultural heritage, languages, sciences, knowledge, convictions, religions and beliefs, sports and games, rites and ceremonies, production methods, technology, institutions and livelihoods, connections with nature and the environment, food and dress. They also encompass the rights to access and participate in heritage and resources that facilitate the processes of identification and development.

Cultural rights include, but are not restricted to, the right to artistic expression and creativity by recognized artists. They include the right of everyone to be creative in all fields of human life, in accordance with article 15 (3) of the International Covenant on Economic, Social and Cultural Rights, on the freedom indispensable for creative activity. States have committed to respect and protect the freedom indispensable for creative activity, which is necessary in the workplace, in social interactions, in artistic expressions and scientific enquiries, and in any activity through which individuals express their aspirations, pursue their own individual and collective development and exercise their intelligence. This is what lies at the heart of cultural rights in democratic societies. At the opposite end stand societies that make an artificial distinction between those who create, on the one hand, and those who are mere executors or consumers, on the other.

In order to implement cultural rights, spaces for creativity related to all human activities must be respected, protected and promoted by States and non-State actors and safeguarded against encroachment by AI. Creativity is not a concept relevant to artistic creation only. Creativity is what gives meaning to all human activities. The safeguarding of it is the necessary response to the sense of a loss of meaning that many people feel in their lives, particularly in their work. Safeguarding space for creativity runs contrary to the imperative to produce more (including art works and scientific works) and faster, which leads to a situation in which people will have to rely on machines in order to write and read an overwhelming amount of works of questionable quality. Creativity requires time, thought and exchange.

The cultural rights approach should include the following elements. First, AI tools must not infringe on the spaces needed for human creativity, and they must be developed with the aim of serving human creativity, not replacing it.

Second, AI tools must never be developed without due consideration of the importance of cultural diversity, and penalties should be imposed for the lack thereof. The elaboration of AI tools and processes should be accompanied by multidisciplinary and culturally diverse teams as a minimum measure to reduce bias in the design of AI systems and operations.

Third, AI systems must be programmed to prevent and redress any discrimination and bias. Understanding and mitigating biases is an essential condition for the deployment of more ethical and equitable AI systems. System transparency is required, as only through knowledge of the data used by AI can these biases be reduced. Improving the explainability of these systems is also crucial for an understanding of how decisions are made using AI tools. For this reason, the involvement of sociologists and human rights experts is important. However, these processes have been completely left in the hands of technicians and technology experts who have very limited understanding and knowledge of human rights obligations and standards regarding discrimination and stereotypes.

Fourth, States must ensure that they pay particular attention to groups that may be more vulnerable or affected, including children, minorities, Indigenous Peoples, women and persons with disabilities. They must ensure that effective measures are in place to address the cultural rights and the free creativity of these sections of the population.

Fifth, even more important is the active and meaningful participation of the individuals and communities whose data has been used, the effective participation of minorities, and the free, prior and informed consent of Indigenous Peoples and local populations. When tools are intended for use by specific groups, particularly minorities and Indigenous Peoples, data sets should be developed with the participation of the concerned groups.

Sixth, due to a power imbalance between various stakeholders, human rights impact assessments are essential prior to the deployment of models and could also help to address systemic bias. Benefit-sharing processes can help in the compensation of communities and creators for the use of their work by AI. These are important measures that States should take to fulfil their legal obligations regarding cultural rights.

Finally, AI tools must never be developed without control and evaluation by humans. Artificial intelligence is not true intelligence, but rather machine learning, an automatic and repetitive copy or imitation of human creativity. It should not be allowed to develop uncritically. It needs to be subject to revision and revisability, and the impacts of its processes need to be accounted for. States must ensure that they have clear lines both for the periodic and inclusive evaluation of such tools as well as for accessible remedies for redress for injustices and violations committed.

Artificial intelligence can improve the life of all only if it is critically assessed and consciously channelled towards the respect, protection and exercise of all human rights. If its uses and impacts are not critically assessed and controlled, AI will limit international recognized human rights, dehumanize social interactions and work environments and infringe on human dignity. So far, the impact of AI on cultural rights has been side-stepped, and the protection of cultural rights in AI has not been effectively regulated. The possibility that AI systems may violate the freedom to develop and engage in creative activity and the right to take part in cultural life is not hypothetical: it is currently happening, unfolding insidiously. There is an urgent need to take a step back from the incessant fascination with AI and recognize the multifaceted ways in which its use can irreparably erode human creativity. It is time that States adopted concrete measures to correct the course and protect the spaces, time and conditions necessary for the protection and exercise of human creative potential.

The Special Rapporteur recommends that Governments, at the international, national and local levels:
(a) Ensure that national and international AI regulatory frameworks explicitly and effectively respect, protect and fulfil cultural rights;
(b) Reaffirm the value of human creativity and adopt measures that ensure the recognition and visibility of works created by humans, particularly in environments where AI-generated content proliferates;
(c) Take measures to ensure the right of meaningful participation of all affected individuals and groups, including through free, prior and informed consent, consent-based protocols and ethical guidelines, in all matters relating to AI, and protect traditional creative expressions and Indigenous knowledge systems from unauthorized or inappropriate use by AI systems;
(d) Ensure that public funding and cultural policies actively support a diversity of creative expressions, including those that do not rely on AI technologies;
(e) Protect and promote data sovereignty in the cultural field, as individual and collective authors must retain control over how their data is collected, stored, used and shared in AI systems;
(f) Strengthen legislative and regulatory frameworks to guarantee the protection of right holders’ moral and material interests when their works are used to train AI models, including through transparency requirements;
(g) Adopt robust legal and technical safeguards to prevent the non-consensual use of artists’ biometric data by AI systems and to regulate the creation and use of synthetic replicas;
(h) Ensure that AI systems are never allowed to develop without human intervention, and ensure that multidisciplinary and culturally diverse teams are considered a minimum measure to prevent and reduce bias in AI systems and processes;
(i) Mitigate gender-based bias in AI systems by applying intersectional approaches, ensuring inclusive data sets and adopting safeguards that uphold the rights of women, girls and gender-diverse individuals, and correct detected discriminatory patterns;
(j) Ensure child-sensitive AI governance by embedding children’s rights principles into AI design, with safeguards against data exploitation and identity stereotyping;
(k) Adopt policies to ensure that recommendation, ranking and filtering systems uphold the right to access and discover a diversity of creative expressions, including those created by underrepresented groups;
(l) Develop and disseminate educational resources and training programmes to strengthen the capacity of all, including technology personnel, to understand, use and critically engage with AI systems;
(m) Stimulate and fund research initiatives aimed at documenting and analysing the effects of AI technologies on human creativity;
(n) Ensure that clear accountability mechanisms are in place to identify and address harms caused by AI systems in creativity, including through independent oversight and accessible complaint procedures;
(o) Require multinational enterprises operating in the cultural and creative sectors to adopt transparent, fair and rights-based approaches to AI development, with particular attention given to creative impacts and in line with international due diligence standards.

The Special Rapporteur recommends that all public, private and civil stakeholders involved in the development and functioning of AI tools and processes cooperate at the local, national and international levels in order to:
(a) Develop common standards and interoperable mechanisms for the identification and traceability of AI-generated content;
(b) Implement cultural rights impact assessments and reviews of AI initiatives and introduce compulsory human and cultural rights trainings for all staff developing AI processes that affect creativity;
(c) Develop and promote fair, transparent and rights-based licensing mechanisms to operationalize the protection of human authors’ moral and material rights, ensuring equitable remuneration and meaningful control for artists and creators;
(d) Adopt measures to ensure transparency and oversight throughout the AI life cycle;
(e) Adopt safeguards against the use of AI to alter, distort or manipulate creative works and narratives in ways that misrepresent or discredit their meaning, origin or intent;
(f) Promote the creation and use of inclusive, representative and fairly sourced cultural data sets, developed with the free, prior and informed consent of affected communities and governed by ethical principles that address structural inequalities and uphold cultural rights;
(g) Promote the development and adoption of open AI tools in the cultural and scientific sector in order to foster innovation and reduce dependency on proprietary technologies;
(h) Promote equitable access to AI infrastructure, tools and training for creators and cultural actors in the Global South in order to prevent deepening global asymmetries in the creation, production and dissemination of cultural expressions;
(i) Encourage collaboration between artists and AI experts to foster mutual understanding and stimulate cross-disciplinary innovation.