A/80/278 creators, AI tools do not support the dissemination of their works or make them more accessible to the public. 31. 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 lowincome 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 32. 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 33. 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. E. Artificial intelligence content, bias and discrimination 34. 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 decisionmaking – tend to reinforce and even exacerbate existing social, cultural, economic and political inequalities. 35. 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 absent 56 or misrepresented in AI-assisted or AI-generated creations. In sectors or platforms __________________ 50 51 52 53 54 55 56 12/21 Submission by Fundación para la Democracia, p. 1. 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”. Submission by Brazil, p. 2. Submission by Eva Nieto McAvoy, p. 3. UNESCO, “Recommendation on the ethics of artificial intelligence”, 2022, para. 29. Submission by Nicolás Madoery, FUTURX, p. 2. Submission by Anna Su, p. 2. 25-12403