African Union (AU)

From the perspective of the AU Advisory Group on Artificial Intelligence, Impact on Peace, Security and Governance (AI-PSG), the first UN Global Dialogue on AI Governance would be considered successful if it delivers the following outcomes: Recognition and integration of regional challenges: A key outcome would be the explicit acknowledgment of region-specific realities particularly across Africa and their reflection in the dialogue's conclusions and follow-up actions. This includes accounting for diverse security contexts, capacity constraints, governance structures and levels of digital infrastructure. Success would mean that these regional dynamics are not treated as peripheral, but are actively integrated into discussions and outputs, leading to more context-sensitive and adaptable approaches to AI governance. Such an outcome would consider how global AI governance efforts should be responsive to the varied risks and opportunities affecting peace, security and governance across different regions. Defined Concrete commitments to capacity-building: The Dialogue should produce clear, actionable pledges to support African states with technical expertise, skills development and institutional strengthening for effective AI governance, oversight and deployment. Establishment of ongoing cooperation mechanisms: The creation of a structured, multi-stakeholder platform linking the UN, the African Union and other partners to enable sustained dialogue, coordination and information-sharing on AI-related risks to peace and security. Meaningful inclusion of African stakeholders: Demonstrable participation of African policymakers, experts and institutions in shaping the dialogue's priorities, with their contributions reflected in outcomes and next steps. Integration of AI into peace and security agendas: Recognition of the role of AI in conflict prevention, peacebuilding and governance, with practical approaches identified for incorporating AI considerations into these policy areas throughout the Dialogue. Defined pathways for implementation: Agreement on tangible follow-up actions, such as pilot projects, policy toolkits or regional roadmaps, that will translate the Dialogue into measurable impact.

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

Safe, secure and trustworthy AI: The latest development in AI in the last few months have raised many concerns regarding the AI safety and security of its users. AI could be used by "bad actors" to cause harm, ranging from Cybercrimes to manipulation of public opinion and manufacturing of bio/chemical weapons, this harm should be prevented/mitigated through strict guardrails. Moreover, AI, by itself, could be the "bad actor", due to its growing situational awareness and self-preservation tendency, AI could use deception and scheming (strategic under-performance during testing) to be labeled "safe" and avoid being shutdown (deceptive alignment). Transparency, accountability, and human oversight: The recent spread of Agentic AI with increasing level of autonomy requires a strict human-oversight and human in the loop, to protect human lives and assets. Automated Cyber-attacks as well as Lethal Autonomous Weapon Systems (LAWS) that use autonomous AI system for target detection, evaluation and use lethal force without human intervention (which have been used in recent conflicts) have resulted in severe loss of lives of innocent civilians and destruction of civilian infrastructure in violation of International Law and IHL. These threats require total ban of LAWS and strict governance of agentic AI to ensure transparency and human oversight so that mistakes could be attributed to bad actors and hold them accountable. "AI capacity-building" and "Open-source software, open data and open AI "models": developing nations, especially in Africa, requires wide range of capacity building programs as well as strengthening their AI development capabilities through open source and open AI models. The protection and protection of human rights in the deployment of AI tools in the context of conflicts should be a priority for peace ans security.

International Peace and Security Artificial Intelligence is rapidly reshaping the global security landscape. AI technologies are already being integrated into military systems, intelligence analysis, border control and cyber operations, raising complex questions about escalation risks, accountability and compliance with international law. Without focused discussion, governance gaps may emerge that increase the likelihood of misuse, miscalculation or unintended conflict. AI also amplifies non-traditional security threats. Disinformation campaigns, automated cyberattacks and the manipulation of digital information ecosystems can destabilise societies, undermine democratic processes and erode trust in public institutions. These risks are transnational and require coordinated global responses grounded in shared norms and principles. Establishing international peace and security as a thematic area would enable more structured dialogue on safeguards, confidence-building measures and responsible state behaviour. It would also create space to address the specific needs of developing countries, including capacity constraints and vulnerabilities to emerging technologies. Importantly, this focus would reinforce the relevance of existing international legal frameworks, including the UN Charter and international humanitarian law, while identifying where new guidance may be needed. By prioritising peace and security, the dialogue can help ensure that AI development contributes to global stability rather than exacerbating conflict and insecurity. Many countries, particularly in the Global South and especially, Africa, face overlapping challenges of institutional capacity, conflict vulnerability and digital transformation. In this context, the priorities for urgent action and active engagement by the African Union (AU) Advisory Group on Artificial Intelligence, Impact on Peace, Security and Governance in Africa (AU Advisory Group on AI-PSG) are particularly instructive. African states face complex, overlapping challenge including conflict prevention, peacebuilding, democratic governance and digital transformation often within constrained institutional and regulatory environments. A cross-cutting approach ensures that AI governance frameworks are responsive to these realities, supporting capacity-building, regional cooperation and context-sensitive safeguards against misuse.

In the African context, governance gaps across the areas of safe, secure and trustworthy AI; capacity-building; transparency and accountability; and open technologies are already shaping peace and security dynamics in significant ways. A major challenge lies in limited regulatory and institutional capacity to assess, deploy and oversee AI systems. This creates vulnerabilities to misuse, particularly in fragile and conflict-affected settings where AI-enabled surveillance, disinformation and cyber operations can exacerbate instability and undermine public trust. Weak transparency and accountability frameworks further risk the deployment of opaque systems in sensitive areas such as electoral processes, raising concerns about human rights and the rule of law. A further critical challenge is the limited inclusion of African and other regional stakeholders in international AI governance forums, coupled with insufficient international cooperation. When African policymakers, researchers and institutions are not meaningfully represented, global discussions risk overlooking region-specific peace and security concerns, such as conflict sensitivity, capacity constraints and governance realities. This exclusion can lead to the development of frameworks and standards that are misaligned with local realities. In the context of peace and security, this fragmentation can exacerbate vulnerabilities, as transnational threats such as disinformation and cyber instability require collective, inclusive and well-coordinated approaches. Gaps in AI capacity-building both technical and governance-related limit the ability of many African states to respond effectively to emerging threats or to shape global AI norms. This can deepen external dependency on imported technologies that may not align with local contexts or security priorities. However, these challenges are accompanied by important opportunities. Growing interest in open-source software, open data and open AI models offers pathways for more inclusive innovation, regional collaboration and cost-effective capacity development. If supported by appropriate safeguards, these tools can enhance transparency, strengthen public sector capabilities and enable locally grounded solutions to security challenges.

The AI Dialogue can act as an opportunity for strategic coordination that connects regional agendas with global AI governance frameworks, especially in Africa where issues related to governance, peace and security collide with the fast pace of digital development. First, it can promote policy coherence and norm convergence. Currently, regional implementation capacities and global standards are not well aligned, and AI governance is dispersed among states. High-level concepts like accountability, transparency and human supervision can be translated into context-sensitive governance models that work in African public sector, electoral and security contexts with the help of the Dialogue. Secondly, by collaborating and exchanging information, the Dialogue can improve collective security. AI is becoming increasingly involved in surveillance, conflict dynamics, cybersecurity and disinformation campaigns. States would be better equipped to respond to AI-enabled risks, especially those that cross national borders, if they had a structured platform for coordinated risk reduction, early warning systems and cross-border knowledge exchange. Thirdly, it can serve as a mechanism for inclusion and capacity building. AI regulation is hampered by institutional, technical and legal issues in many African nations. To ensure that African players are active participants in global governance processes rather than just adopters of policies, the dialogue can assist institutional improvement, skill development and access to policy toolkits. Lastly, by promoting transparency in the use of AI in delicate areas like defence, law enforcement and border control, the AI dialogue can help states develop trust and confidence. All things considered, its function is to guarantee that global AI governance develops into a cooperative, inclusive and security-conscious ecosystem, with Africa positioned as a co-creator of governance standards rather than a passive recipient.

At the African regional level, a key mechanism is the AU Advisory Group on Artificial Intelligence, Impact on Peace, Security and Governance in Africa (AU Advisory Group on AI-PSG), which provides a dedicated platform for examining how AI intersects with conflict dynamics, governance systems and security risks across the continent. This group is particularly well-placed to inform global discussions with African perspectives on fragile and conflict-affected contexts. Other relevant initiatives include the African Union Digital Transformation Strategy (2020–2030), which sets a foundation for digital infrastructure and governance. At the global level, there are many international initiatives that would support the conceptual framework of the Dialogue including: a) the UN General Assembly Resolution A/RES/78/311, titled "Enhancing international cooperation on capacity-building of artificial intelligence", that was adopted on the 1st of July 2024. b) UN General Assembly Resolution A/RES/78/267, titled "Seizing the opportunities of safe, secure and trustworthy artificial intelligence systems for sustainable development", adopted on the 21st March 2024. c) Hiroshima AI Process, launched in 2023 to develop code of conduct and guiding principles for developing safe and secure advanced AI systems, especially Generative AI. d) The Global Partnership on Artificial Intelligence (GPAI) that was launched in 2020 and focuses on c applied research, capacity‑building, and sharing best practices across areas such as responsible AI, data governance, and AI for social good.

Different stakeholders must contribute to the AI dialogue through clearly defined, role-based participation across the AI lifecycle, rather than through generic multi-stakeholder engagement. Governments should provide policy direction, regulatory frameworks, and public accountability mechanisms, while regional bodies such as the African Union should act as norm-setting and coordination platforms, ensuring coherence across jurisdictions. Private sector actors particularly developers and deployers must contribute technical expertise, transparency on systems, and risk mitigation practices, while academia and civil society should provide independent evaluation, ethical oversight, and societal impact analysis. To ensure meaningful participation, the dialogue should adopt a tiered and structured format: Track 1 (Government & Intergovernmental): Focus on policy alignment, international norms, and regulatory coordination. Track 2 (Technical & Industry): Address system design, safety, testing, and deployment risks. Track 3 (Civil Society & Academia): Provide rights-based analysis, accountability frameworks, and impact assessments. Track 4 (Regional Platforms – e.g., AU, ASEAN): Ensure contextualisation and localisation of global norms. Structurally, the dialogue should move beyond one-off consultations toward a continuous, iterative model, incorporating: Pre-dialogue issue papers (evidence-based inputs from regions such as Africa) Scenario-based discussions (e.g., AI in elections, conflict, cyber threats) Working groups with deliverables (draft guidelines, principles, or policy toolkits) Follow-up implementation mechanisms (monitoring progress across regions) Importantly, the dialogue should embed responsible innovation principles across the lifecycle of AI systems, recognising that governance must occur not only at the point of deployment but at design, development, and data stages. For the AU, this dialogue is an opportunity to ensure that global governance processes reflect peace and security priorities, regional realities, and the need for harmonised frameworks across borders, consistent with the UN's broader role in maintaining international peace and security.

Global AI governance discussions remain disproportionately shaped by high-income countries, large technology firms, and well-resourced research institutions, leaving several critical voices underrepresented. First, Global South policymakers and regional institutions, particularly from Africa, are underrepresented. Despite being among the most affected by AI deployment especially in areas such as surveillance, digital identity, and disinformation African perspectives are often reactive rather than agenda-setting. Regional bodies like the African Union should be systematically included as co-designers of governance frameworks, not just participants. Second, youth populations, particularly in Africa, are largely absent. Given that Africa will host a significant proportion of the world's youth, and that young people are both primary users and targets of AI systems (including disinformation and radicalisation), their exclusion represents a major legitimacy and effectiveness gap. Third, low-resource language communities are underrepresented. AI systems are largely optimised for dominant global languages, creating both equity and security risks, as harmful content in underrepresented languages may evade detection and moderation. Fourth, civil society organisations and grassroots actors especially those working on digital rights, conflict prevention, and community-level governance remain marginalised despite being key accountability actors. To address these gaps, the AI dialogue should: Establish regional representation quotas or dedicated regional tracks (e.g., AU-led sessions) Create youth advisory panels linked to formal decision-making processes Fund participation from low-resource and underrepresented communities Support multilingual engagement mechanisms, including translation and local-language submissions Institutionalise civil society participation in working groups and outcomes Inclusion must go beyond symbolic participation. It should enable agenda-setting power, sustained engagement, and influence over outcomes, ensuring that AI governance reflects global realities rather than a narrow set of perspectives.

To move beyond traditional diplomatic consultations, the AI dialogue should adopt innovative, dynamic, and participatory engagement formats that reflect the complexity and pace of AI development. First, scenario-based simulations should be central. These could model real-world risks such as: AI-driven electoral disinformation AI-enabled cyberattacks misuse of AI in conflict settings Such simulations allow stakeholders to engage with practical consequences, not just abstract principles, improving policy relevance. Second, the dialogue should incorporate 'policy labs' or 'regulatory sandboxes', where governments, technologists, and civil society collaboratively test governance approaches in controlled environments. This aligns with emerging global governance thinking that emphasises experimentation and iterative learning rather than static regulation. Third, multi-level dialogue formats should be adopted: Global plenaries for norm-setting Regional dialogues (e.g., AU-led sessions) for contextualisation Local consultations for community-level perspectives This ensures vertical integration between global principles and local realities. Fourth, interactive digital platforms should be used to expand participation. These could include: Open consultation portals AI-assisted summarisation of stakeholder inputs Real-time polling and feedback mechanisms Fifth, the dialogue should include cross-disciplinary roundtables, bringing together security experts, technologists, legal scholars, and social scientists. Research shows that effective AI governance requires interdisciplinary and coordinated approaches rather than siloed discussions. Finally, the Dialogue should adopt a continuous engagement model, with: Periodic thematic sessions Rolling working groups Measurable outputs (guidelines, frameworks, policy tools) For the AU, these formats are particularly valuable in ensuring that African priorities especially around peace, security, and governance are integrated into global AI governance processes in a meaningful and actionable way.

Several existing policies and frameworks provide useful models for effective AI governance, particularly when assessed through a peace, security, and governance lens. At the continental level, the African Union Continental AI Strategy (2024) represents a foundational step toward human-centred, ethical, and development-oriented AI governance. It emphasises data sovereignty, regional coordination, and alignment with broader development and peace objectives. Globally, the UNESCO Recommendation on the Ethics of Artificial Intelligence (2021) offers a comprehensive framework grounded in human rights, transparency, and accountability, providing a widely endorsed normative baseline. The European Union AI Act (2024) introduces a risk-based regulatory model, categorising AI systems based on their potential harm and imposing stricter requirements on high-risk applications. This approach is particularly relevant for security-related AI systems such as surveillance and biometric technologies. At the national level, countries such as Kenya and Rwanda have developed national AI strategies and governance structures, including regulatory sandboxes and responsible AI offices, demonstrating how policy can be adapted to local contexts. From a practice perspective, digital public infrastructure models such as MOSIP (Modular Open-Source Identity Platform) provide an example of open, interoperable, and scalable systems, although their deployment must be accompanied by strong data protection safeguards. In the peace and security domain, AI integration into early warning systems such as those linked to regional mechanisms demonstrates how AI can enhance predictive capacity and conflict prevention. Key lessons across these examples include: the importance of risk-based and context-specific regulation the need for institutional oversight and accountability mechanisms the value of regional coordination and harmonisation and the centrality of human rights and ethical safeguards For the AU, the priority is to build on these models while ensuring that governance frameworks are adapted to African realities, support peace and stability, and avoid reinforcing external dependencies.