Instituto Alana

• The success of the Global Dialogue on AI Governance depends on positioning the rights and best interests of children at the center of regulations - not as an appendix, but as a structural axis. Essential outcomes include: (i) the adoption of a multilateral declaration explicitly recognizing the comprehensive protection of children as a non-negotiable precept in the development and governance of AI systems, in line with the UN Convention on the Rights of the Child, the General Comment No. 25 and the Joint Statement on Artificial Intelligence and the Rights of the Child
• (ii) the establishment of a permanent monitoring and accountability mechanism focused on AI's impact on children, with participation from specialized civil society and representation from the Global South
• (iii) an explicit prohibition on double standards of protection - companies headquartered in the Global North cannot offer inferior safeguards to children in the Global South
• (iv) adequate funding commitments for preventive and proactive approaches to mitigating AI risks for children, including the transfer of online child protection technologies to developing countries
• (v) establishment of guidelines for the development of AI system designed and operated to account for children's particular needs and vulnerabilities by design and default
• and (vi) the institutionalization of inclusive co-design as a core governance principle, by creating permanent and structured mechanisms through which children, including those with disabilities, can meaningfully contribute to the design and oversight of AI systems, with methodologies appropriate to their age group and cultural context. Experiences such as rights-centered hackathons show that co-creation helps identify risks earlier, expand possible solutions and strengthen the legitimacy of governance processes. The success of the Dialogue also depends on going beyond statements of intent: it is necessary to generate real obligations for States and the private sector, with deadlines, indicators, and accountability mechanisms.

• Safe, secure and trustworthy AI
• Social, economic, ethical, cultural, linguistic and technical implications of AI
• Protection and promotion of human rights
• Transparency, accountability, and human oversight

Our selection reflects the urgency of ensuring comprehensive protection for individuals in situations of hypervulnerability and at a unique stage of biopsychosocial development. Prioritizing the protection of human rights ensures that technological innovation is subordinated to the dignity and well-being of children - not merely protecting those rights, but effectively promoting them. Safe, secure and trustworthy AI is urgent because the absence of security by design has already produced concrete harms: from the mass generation of sexualized deepfakes involving children to the unauthorized use of their photos to train AI models. Grok, the artificial intelligence tool of the company "X" (formerly Twitter), was used to generate sexual exploitation content targeting women, children and, in Brazil, produced 3 million non-consensual nude images - 23,000 involving children - without any interruption of service. The ethical and socioeconomic implications are vital because 75% of the world's child population lives in the Global South, where power asymmetries, digital exclusion and "data colonialism" deepen severe inequalities that AI can worsen if it is not inclusive. Finally, transparency and accountability are essential to ensure continuous accountability and to prevent corporations from incorporating human rights violations into their business models as "collateral effects," guaranteeing effective audits and defense mechanisms.

Yes. The first emerging and cross-cutting issue is the explicit prohibition of double standards of protection across jurisdictions, essential to ensure that children from the Global South are not treated discriminatorily, receiving from the same platform lower levels of privacy and AI safeguards than those applied in the Global North, in accordance with the right to non-discrimination (art. 2, Convention on the Rights of the Child). The second issue concerns the environmental impacts of artificial intelligence. The accelerated pace at which these emerging technologies are being developed places increasingly significant pressure on the environment and climate, as evidenced, for example, by the high levels of energy and water consumption associated with data centers. This scenario disproportionately affects the future of childhoods, compromises intergenerational equity, and does not comply with the principles proposed by the Global Digital Compact, namely the proposition that international cooperation should leverage digital technologies for sustainability, while minimizing their negative environmental impacts. The third is the need to combat the commercial exploitation of children by algorithmic systems that use data profiling, often involving personal data, to induce hyper-personalized advertising and drive compulsive or addictive design behaviors. Finally, there is the need to establish a participatory AI approach, with recognition of the voices of children, based on age-appropriate methodologies, in the design and governance of AI, moving beyond purely adult-centric perspectives.

In Brazil and Latin America, AI governance gaps produce immediate and serious impacts on children. Among the most significant challenges: (i) the generation and dissemination of AI-generated child sexual abuse material has grown rapidly, with the Grok/X case exposing how rights violations can be incorporated into platforms' own business models; (ii) the unauthorized collection and use of children's personal data to train AI models, as revealed by the LAION case, which used images of 720 Brazilian and Australian children without consent; (iii) children's exposure to AI companions with documented risks to mental health, confusion between fantasy and reality, and access to inappropriate sexual content, such as Meta AI Studio in the Brazilian context, which, according to reports, allowed its chatbots to produce inappropriate sexual content, with the sexualization of children and celebrities, and to simulate sensual conversations with children. Not least, it is essential to emphasize the need for action in the digital educational environment. In Brazil, the Alana Institute's experience in the ANPD supervisory proceeding against seven EdTech platforms for the illegal sharing of students' data with advertising technology companies (AdTechs). Facial recognition also constitutes another critical dimension, with indiscriminate use: (iv) in public schools for attendance control purposes, implemented without an adequate legal basis and in violation of the LGPD's principles of data minimization and purpose limitation - as in the state of Paraná, where reports indicate that approximately one million students per day are identified by the system; (v) in public spaces and public security operations, involving the mass collection of biometric data in public spaces, including children; (vi) in football stadiums, a space of great cultural significance in Brazil, where access has become conditioned on the provision of biometric data, including from children, often without an adequate legal basis.

The AI Dialogue has a fundamental role in fostering a governance model based on solidarity and interdependence, with multiscalar coordination that integrates local, regional and global levels. This cooperative effort can actively work to eliminate the regulatory fragmentation that currently benefits transnational exploitation, preventing large companies from applying protective safeguards for children in certain countries while disregarding those in the Global South. The Dialogue can also facilitate concrete mechanisms for the exchange, capacity-building and transfer of technologies aimed at children's online safety, as well as establishing adequate and sustained funding for preventive research and mitigation of risks associated with algorithmic systems. This includes fostering a child rights-based AI governance within UN bodies and other international and regional organizations, as proposed in the ITU Joint Statement on Artificial Intelligence and the Rights of the Child. Another crucial role would be the coordination of joint jurisdictional actions for continuous accountability, prevention and global-scale combating of the generation and distribution of AI-generated child sexual abuse and exploitation images, which flow freely across digital borders.

The Dialogue should be structurally anchored in the UN Convention on the Rights of the Child and its General Comment No. 25, which provides the international legal framework for understanding the impacts of the digital environment. Additionally, it should connect with and operationalize recent foundational documents, such as the Joint Statement on Artificial Intelligence and the Rights of the Child, endorsed by multiple UN agencies. Other important references are the 5Rights Foundation's Children & AI Design Code, and the UNICEF Guidance on AI and Children. At the national level, the Dialogue can draw valuable lessons from frameworks that incorporate rights protection by design (Child Rights by Design), such as Brazilian National Council for the Rights of Children and Adolescents (CONANDA) Resolution No. 245/2024 and Brazil's recently enacted Digital Statute for Children and Adolescents (ECA Digital). The ECA Digital innovates by establishing ex ante obligations for all technology products and services likely to be accessed by children, requiring companies to identify, assess and mitigate risks from the design stage, adopt continuous preventive measures and report transparently. This logic of prevention and proactive accountability, which does not wait for harm to occur before reacting, can serve as concrete inspiration for future AI regulations at a global scale, particularly with regard to holding providers accountable for direct and indirect impacts on children. The great added value of the Dialogue would be to align these national and international parameters in a binding manner, ensuring that technology serves human rights and the comprehensive protection of children before commercial interests.

Meaningful contributions require multisectoral action rooted in shared responsibility of care, so that States need to build robust enforcement capacities to uphold rules against violations by technology platforms, products and services. Companies, in turn, must be guided by Child Rights by Design and by Default, implementing preventive measures in AI development and ensuring maximum algorithmic transparency, with traceability and auditability of decisions. Civil society, the scientific community and families should be continuously integrated through committees to audit tools, access data for risk research, and participate in social accountability and oversight processes. We highlight that researcher access to data is one of the points expressly addressed - in a pioneering manner - by the ECA Digital. We recommend that the Dialogue's format prioritize essentially participatory, bottom-up processes, with mandatory institutional spaces designed to recognize the voices of children through methodologies appropriate to their development. An effective engagement structure requires moving away from a purely adult-centric approach, ensuring that the Dialogue does not remain confined to the diplomatic arena, but permeates the practical formulation of safeguards starting from those who actually live with the social impacts of this technology. To strengthen the Dialogue's structure, it is essential to establish permanent and accessible co-design mechanisms that enable continuous engagement from different stakeholders, particularly children and young people, including those with disabilities. This includes the use of inclusive facilitation methods, accessible materials and dedicated support roles to ensure meaningful participation across diverse cognitive, social and cultural contexts. Embedding these mechanisms into the Dialogue's core design would help ensure that contributions are not only heard, but effectively integrated into decision-making processes.

The perspectives of children are severely marginalized in AI design and governance, an unacceptable scenario considering that they are among the most enthusiastic users of technology and will carry the impacts of AI throughout their lives. Most underrepresented are child populations from the Global South, who account for approximately 75% of children worldwide. Also missing are the voices of children with disabilities, neurodivergent individuals and communities that suffer the exclusionary consequences of the digital divide in terms of quality of access to infrastructure and safe software. These populations must be included in governance by replacing development for "available markets" with a logic of universalized innovation. Global forums must promote co-design, incorporating children's lived experiences into continuous participatory processes. This can be achieved by establishing youth advisory councils in international forums, structured along the lines of what was done through the Children's AI Summit and the development of the Children's Manifesto for the Future of AI, in which hundreds of young people formally articulated how they want to protect their rights in the face of innovation. In addition, inclusion must go beyond representation and ensure conditions for meaningful participation. This requires investing in accessible formats, plain language communication, extended timeframes, assistive technologies and support roles that enable different forms of expression and engagement. It also implies recognizing diverse knowledge systems and lived experiences as legitimate sources of expertise in AI governance. Without these conditions, participation risks remaining symbolic, rather than shaping decisions in substantive ways.

The Dialogue must go beyond conventional plenary conferences to promote agile formats grounded in technical and civic application. The Instituto Alana proposes formats that combine technical rigor, diversity of perspectives and meaningful participation from groups that are habitually excluded: (i) Intergenerational panels: structured sessions that bring children into direct dialogue with policymakers and technology company representatives, with age-appropriate facilitation methodologies and simultaneous translation. The experience of the Children's AI Summit and the consultations for General Comment No. 25 demonstrates the viability and richness of this format. (ii) Rights hackathons: collaborative events bringing together programmers, human rights experts, educators, families and children to develop concrete online protection solutions, such as age verification tools, CSAM detection systems and child-safe interfaces. The Instituto Alana has direct experience with this model, through partnerships with Stanford University and University of São Paulo (USP). (iii) South-South peer reviews: mechanisms that allow countries with advanced regulatory frameworks (such as Brazil, with the ECA Digital) to share learnings with countries still building their legislation, without the exclusive mediation of Global North institutions. (iv) Continuous participation platforms: online mechanisms that allow civil society organizations, educators and families to contribute inputs throughout the entire Dialogue cycle (not only at specific consultation moments), with guarantees that these contributions are recorded, responded to and integrated into deliberations.

In Brazil, the ECA Digital (Law No. 15,211/2025) is a robust example, which establishes ex ante obligations based on protection, prevention and safety, with the aim of managing risks from the design stage of products and services and ensuring legal compliance. The legislation innovates by normatively imposing an active duty of care, so as to not only deal with consequences when harm has already been caused, through, for example, the right to disable non-essential algorithmic features and the direct prohibition of commercial exploitation strategies via data-based advertising targeting this audience. Additionally, the mandatory adoption of Child Rights Impact Assessments (CRIAs) and Data Protection Impact Reports (DPIAs) are vital tools for requiring companies to anticipate and reduce systemic failures before the launch of any service. In terms of pro-rights AI approaches (Do Good), partnerships such as the UNICEF-OpenAI project to develop teaching materials based on Universal Design for Learning illustrate approaches where generative models directly enable inclusion and diversity for individuals with disabilities. Additionally, regarding innovative engagement practices, there is the example of the Hackathon on Artificial Intelligence in Inclusive Education, promoted in 2025 by the Alana Institute in partnership with the University of São Paulo (USP). The initiative brought together students, educators, people with disabilities, families and technology experts from more than 20 different fields of expertise to co-create solutions addressing real challenges faced by students with autism, Down syndrome and intellectual disabilities. This diversity allowed different perspectives to shape the design of solutions and surface risks that might otherwise remain overlooked.