Paradigm Initiative (PIN) works to connect underserved young Africans with digital opportunities and ensures the protection of their rights. Across our regional offices in Kenya, Nigeria, Senegal, Zambia, Zimbabwe and beyond, PIN works to connect under-served African youth with improved livelihoods through digital inclusion and digital rights programs. PIN advocates for the promotion and respect of digital rights and digital inclusion to ensure an enabling environment for fundamental rights. In view of this mandate, PIN submits comments and recommendations to the Global Dialogue on AI Governance.

The Global AI Dialogue should use a multistakeholder approach and focus on a few key topics. The Dialogue should ensure the participation of non-governmental stakeholders to represent key concerns from the Global South. We acknowledge that the Terms of reference and modalities for the establishment and functioning of the Independent International Scientific Panel on Artificial Intelligence and the Global Dialogue on Artificial Intelligence Governance in paragraph 11 encourage States, private sector, financial institutions, foundations and other donors in a position to do so, to support the effective functioning of the Panel and Dialogue, including by facilitating the participation of representatives and stakeholders of developing countries by offering travel support, through voluntary contributions that are made public in the interest of full transparency. The Global Dialogue should prioritise adequate resourcing aimed at facilitating participation of Global South actors, especially civil society actors, if multistakeholderism envisaged in the Terms of Reference is to be a reality. Also, it is pertinent to highlight the need for stakeholders from the Global South to be part of the agenda-setting of core priorities for discussion in the Dialogue with clarity of engagement modalities. Institutionalised participation by civil society organisations from the Global South will be key and the São Paulo Multistakeholder Guidelines will serve as a reference. When setting its agenda, the Dialogue should discuss AI red lines and determine prohibited AI use cases. This will build on work already done by the UN General Assembly (A/RES/78/265), the Office of the High Commissioner for Human Rights (OHCHR) (A/HRC/60/45), United Nations (UN) human rights bodies like Special Procedures and Treaty Bodies. The Dialogue should cover clear guidelines for human oversight and human-in-the-loop systems in regulatory frameworks. The Dialogue should develop best-practice guidelines for human rights due diligence and risk reduction in AI development, deployment, and use.

• Safe, secure and trustworthy AI
• AI capacity-building
• Transparency, accountability, and human oversight
• Protection and promotion of human rights

To ensure safe, secure and trustworthy AI, there is a need to define AI 'red lines' and implement human rights due diligence throughout the AI lifecycle. This is important for establishing a rights-based regulatory framework. Given the rise in reports of erratic behaviour by agentic AI tools, there is a clear need for guidance on essential security features to prevent unacceptable AI outcomes. The Dialogue should also establish a strong baseline for international oversight, promoting transparency and accountability in AI. The Dialogue should also develop guidelines on human oversight to ensure development, deployment, and use of AI are in line with international standards. In a fragmented AI governance landscape, the Dialogue has a unique opportunity to promote regulatory coherence by championing a human-rights-focused framework for AI governance. We envision a Dialogue that is not only development-focused but also balanced with a human rights-based approach, calling for AI deployment anchored in respect, protection and promotion of human rights in the field of artificial intelligence. To support the goals of accountability and transparency, there is a need to promote AI human rights due diligence, which will help the international community with evaluations, remedies and redress. Significant efforts by OHCHR and UN human rights mechanisms have aimed to apply and enforce international human rights law to the internet and digital technologies, including AI. The Dialogue should refocus political attention on these efforts and advance them. AI Capacity Building is a priority because digital divides persist in Africa and in other Global South countries. In light of the increased risks of harm caused by AI, there is a need to provide guidance on essential security features of agentic AI and other AI, to prevent unacceptable AI outcomes. Paradigm Initiative's 2025 Londa report documents limited internet penetration in African countries and spotlights unreliable electricity. The infrastructural challenges and low internet penetration result in the exclusion of communities from AI adoption, as AI use requires reliable internet access and electricity. Through trainings of policymakers, media students, journalists and lawyers, Paradigm Initiative has noted low capacity in the understanding of AI harms, noting the need for increased capacity building.

Two other important topics that are cross-cutting across are the environmental impact of AI infrastructure, like hyperscale AI data centres, and labour protections for those involved in AI development. These two topics should be included in the high-level plenary discussions. Understanding AI's physical infrastructure and environmental effects is crucial for determining AI's social, economic and ethical implications, particularly its negative effects on human rights, such as the right to an adequate standard of living for communities hosting large AI data centres and on labour conditions. These impacts can systematically deepen poverty and inequality. AI's effect on employment is already evident, with reports of large AI companies outsourcing tasks like data labelling and content moderation to external firms, often with inadequate pay and subpar working conditions. Addressing these issues is vital for tackling the human and societal costs of AI. As evidence grows, these topics will need dedicated discussion in future forums.

Safe, secure and trustworthy AI Across the Global South, the spread of AI-powered non-consensual intimate images (NCII) and Tech Facilitated GBV (TFGBV) is contributing to a decline in democracy by silencing women activists, human rights defenders, and journalists through the generation of deepfakes, discriminatory decision-making in digital social protection systems, exclusionary AI-powered digital identity systems and increased information disorders. AI Capacity Building Across the African continent, infrastructure gaps and uneven policy adoption and implementation is resulting in a growing AI divide. PIN's Londa reports documents that at least 17 out of 54 countries are at an advanced stage of AI policy development and deployment, while 36 are at a nascent stage. Despite the AI Continental Strategy, AI regulation and use remain uneven. This fragmentation exacerbates digital divides. Apart from the accessibility of AI systems to remote and underserved communities, governments adopting AI policies have not all included AI risk criteria or established AI red lines, leaving AI use open to abuse and misuse. Additionally, the environmental impacts of AI-powered data centres are not widely understood. These examples highlight the need for targeted AI capacity-building for policymakers and stakeholders across the AI value chain. Protection and promotion of human rights The use of human rights-oriented regulatory frameworks for AI remains weak or nonexistent, and the establishment of independent oversight bodies to manage AI-related violations, harms, and abuse is limited. In Africa, this issue persists despite guidance from the African Union's Continental Strategy, which urges governments to create independent oversight institutions to address AI-relevant violations, harms and abuse. As a result, AI has been designed, developed, and deployed across various jurisdictions in ways that often do not meet current international and regional human rights law obligations. Implementing a human rights-compliant regulatory framework for AI should involve human rights due diligence, including human rights risk assessments. These assessments and mitigation measures should be applied throughout the AI lifecycle, covering both technologies and specific applications. Continuous evaluation is necessary to identify and ban unacceptable technologies or uses. Without such frameworks, we continue to see the development and deployment of AI systems that pose serious human rights risks, such as mass surveillance, crime prediction tools and social scoring, as well as future risks like AI systems that could autonomously conduct cyber-attacks or aid in weapons development. Regulation of Data Centres is happening at a slow pace and with limited alignment to Sustainable Procurement Guidelines for Data Centres and Servers developed by the United Nations Environmental Programme. This fragmentation results in increased harm, as data centres result in environmental harm in countries through their excessive energy and water demands and carbon emissions amid limited national infrastructure. Transparency, accountability, and human oversight The enforcement of a human rights-respecting regulatory framework also requires guarantees of transparency, ongoing evaluation and accessible and effective remedy and redress mechanisms. In the absence of these guarantees, communities are not notified when they have been subject to AI-driven decision-making and are unable to access remedies or redress for AI-related harms.

The Global Dialogue on AI Governance should foster multistakeholder discussions, ensure non-state actors' participation, including civil society participation and promote international cooperation on AI governance. It should focus on a few key topics and leverage the work of existing AI governance initiatives. Priority should be given to issues relating to human rights and those closely linked to the AI Independent Scientific Panel's work, providing a forum for discussing and debating the Panel's annual reports. The Dialogue should develop recommendations on prohibited AI practices and use cases, create guidelines on human oversight and human-in-the-loop processes within regulatory frameworks and establish best practices for human rights due diligence and risk assessments. Outcomes should be endorsed by the UN General Assembly to foster global consensus.

The Global Dialogue should build upon and link to other mechanisms and processes developed under the World Summit of the Information Society (WSIS) and the Global Digital Compact (GDC). It should leverage the Internet Governance Forum (IGF) and its broader network of National and Regional Initiatives (NRIs), alongside the ongoing work of the CSTD Working Group on Data Governance. The co-facilitators should use the IGF as a midpoint assessment of the first Dialogue's outcomes and to define priorities for future discussions in the second Dialogue. The Global Dialogue should collaborate with the IGF Secretariat and the IGF MAG to explore how the Dialogue can integrate with the IGF's intersessional workstreams, thereby promoting high-level outcomes through a multistakeholder approach. The Global Dialogue should use the Multistakeholder strategies outlined in the NETMundial+10 outcome document and the Five-Point Plan for an Inclusive WSIS+20 review, to ensure multistakeholder participation. Furthermore, the Global Dialogue should establish strong links with the activities of UN agencies such as OHCHR and other human rights mechanisms to support the effective enforcement of international human rights law in AI governance.

The co-facilitators should provide opportunities for stakeholders to contribute to the AI Dialogue through written submissions, consultations, sessions and co-chairing, ensuring diverse voices shape AI governance. The co-facilitators should provide an opportunity for stakeholders to inform the agenda and deliberations. Non-state actors such as civil society organisations, including youth, indigenous people and persons with disabilities, private sector, academia, and technical experts should be provided a platform to share their views, to ensure inclusive and meaningful participation of all relevant stakeholder groups. The co-facilitators of the Global Dialogue should create funding mechanisms to support Global South participation, including travel and visa support for Global South communities and stakeholders. They should ensure diverse participation by applying inclusive criteria to the accreditation processes. Including facilitating remote participation for stakeholders who cannot participate in person and those without ECOSOC status. The cofaciliators should make use of the IGF's existing intersessional workstreams to bring in a wide range of expertise to inform the Dialogue's outcomes. When designing the format and structure of the Global Dialogue, the co-facilitators should: Provide a platform that allows for multilingual formats, to make it easier for underserved stakeholders to contribute and share their insights on key topics Publish clear timelines for input and milestones, and provide adequate time for stakeholders to submit input in different formats. Ensure that every speaker from the Global South have space to share their insights so as to ground discussions in lived realities from underserved communities. Live audience question and answer segments with curated questions from in‑person and remote participants Create a segment for testimonials from communities with lived experience of AI-related harms. Maintain a centralised repository of all input received and draft documents, their status and stakeholders consulted. Coordinate the Global Dialogue, WSIS+20 and the AI for Good programmes so that there are no overlaps between key human rights sessions and key sessions with policymakers, to ensure full participation by all stakeholders in key human rights sessions. Facilitate meaningful stakeholder consultations, leveraging existing community convenings to host 'satellite' consultations Structure sessions to allow non-governmental and governmental stakeholders to respond to each other in real time, showcasing diversity of approaches towards one selected topic rather than isolated interventions. Include a dedicated segment focused on outlining the outcomes of the Global Dialogue.

Civil society, women human rights defenders, environmental activists and those impacted by the construction of data centres, data annotators, groups negatively impacted by the use of AI systems in social welfare, policing, border control and other sector and communities in the Global Majority are underrepresented in the AI policy discourse. Their participation is necessary to understand the ethical, social, and environmental impacts, harms, inequalities, and risks raised by AI and to ensure that discussion of policy solutions is grounded in real-life use cases and shaped by real-life experiences.

When designing the format and structure of the Global Dialogue, the co-facilitators should: Provide a platform that allows for multilingual formats, to make it easier for underserved stakeholders to contribute and share their insights on key topics Publish clear timelines for input and milestones, and provide adequate time for stakeholders to submit input in different formats. Ensure that every speaker from the Global South have space to share their insights so as to ground discussions in lived realities from underserved communities. Live audience question and answer segments with curated questions from in‑person and remote participants Create a segment for testimonials from communities with lived experience of AI-related harms. Maintain a centralised repository of all input received and draft documents, their status and stakeholders consulted. Coordinate the Global Dialogue, WSIS+20 and the AI for Good programmes so that there are no overlaps between key human rights sessions and key sessions with policymakers, to ensure full participation by all stakeholders in key human rights sessions. Facilitate meaningful stakeholder consultations, leveraging existing community convenings to host 'satellite' consultations Structure sessions to allow non-governmental and governmental stakeholders to respond to each other in real time, showcasing diversity of approaches towards one selected topic rather than isolated interventions. Include a dedicated segment focused on outlining the outcomes of the Global Dialogue

The OHCHR and the UN human rights mechanisms have been at the forefront of calling for alignment of AI systems with international human rights law, insitsting on mandatory human rights due diligence, setting principles for AI governance and outlining prohibitions. Other normative instruments that offer concrete solutions and embed human rights safeguards include the Council of Europe's Framework Convention on Artificial Intelligence (2024), the first binding treaty requiring AI systems to align with human rights, democracy, and rule of law, the United Nations Interregional Crime and Justice Research Institute (UNICRI-INTERPOL) Toolkit for Responsible AI Use by Law Enforcement, which sets operational standards and export control frameworks. These instruments help prevent harmful exports while promoting accountability and ethical deployment of AI. The Council of Europe's Committee on Artificial Intelligence (AI) adopted the Risk and Impact Assessment of Artificial Intelligence Systems from the Point of View of Human Rights, Democracy and Rule of Law (HUDERIA). The Organisation for Economic Cooperation and Development (OECD) and the Global Partnership on AI (GPAI), developed the OECD Recommendation on AI. The United Nations Scientific and Cultural Organisation developed the Readiness Assessment Methodology and Recommendations on the Ethics of AI. For more information, please contact Thobekile Matimbe -Senior Manager, Partnerships and Engagements on [email protected]