Austrian Institute of Technology / PeaceTech Alliance

The first Global Dialogue on AI Governance would be a success if it moves beyond high-level principles and creates a credible pathway for action grounded in real-world conditions. First, it should recognise that AI governance is not only about models and regulation, but also about data, participation, and power. Current systems often reflect structural imbalances, including the underrepresentation of conflict-affected, low-resource, and non-English-speaking communities. Addressing this gap should be a core outcome. Second, the Dialogue should produce practical follow-up mechanisms. This could include dedicated workstreams on context-sensitive AI governance, practitioner-informed evidence gathering, and pilot approaches that test how responsible AI can be designed and applied in sensitive environments, including across different stages of the conflict cycle. Third, it should build on existing practice rather than reinventing the wheel. Multi-stakeholder initiatives are already emerging that bring together government, research institutions, peace organisations, and technical actors to explore human-centric and ethical approaches to AI and data. In Austria, for example, the PeaceTech Alliance is helping to create collaboration across sectors on exactly these questions. Efforts like these should be recognised and connected to global processes. Finally, the Dialogue should broaden who is recognised as a governance actor. Practitioners, civil society organisations, community actors, and local experts should be treated as contributors to governance, not only as end users or data sources. A successful outcome would therefore be a more inclusive, practice-grounded process aligned with human rights and the principle of Do No Harm.

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

These priorities reflect where governance gaps are most visible in practice, particularly in fragile and conflict-affected settings. Safe, secure and trustworthy AI is critical because failures in design or deployment can have direct consequences. In PeaceTech contexts, where technologies are used to support mediation, analysis, or coordination, poorly adapted systems can increase risk rather than reduce it. The social, ethical, cultural, linguistic and technical implications of AI are central because many systems still reflect narrow assumptions, particularly Global North, English-language, and high-connectivity contexts. This creates a clear mismatch between design and lived realities in many environments where PeaceTech is applied. Protection and promotion of human rights ensures that governance remains grounded in dignity, participation, and agency. In PeaceTech, this includes recognising communities and practitioners not only as users, but as co-creators of data and systems. Transparency, accountability, and human oversight are necessary because frontline practitioners often have limited visibility into how systems are designed or used. In PeaceTech applications, meaningful oversight must include those with contextual and lived knowledge. Together, these priorities support a more human-centric and context-aware approach to AI governance.

A key cross-cutting issue is the relationship between AI governance and conflict sensitivity. AI systems are increasingly deployed in environments shaped by inequality, mistrust, and instability, yet governance frameworks rarely address these dynamics. This is particularly relevant in emerging PeaceTech applications. A second issue is data justice and community-level data governance. Current debates focus heavily on models and regulation, but less on where data comes from, who controls it, and whether communities have agency over its use. In PeaceTech contexts, this is critical, as data practices can either build trust or reinforce perceptions of extraction. A third issue is the disconnect between those who design AI systems and those who experience their impacts. Communities and practitioners are often treated as data sources rather than co-designers or governance actors. This gap is frequently observed in PeaceTech initiatives, where tools are introduced without sufficient grounding in local realities. These issues highlight the need for governance approaches that integrate participation, ownership, and context from the outset.

In the peacebuilding and international policy sector, governance gaps are already evident. A key challenge is the disconnect between design assumptions and operational realities. Many systems are built for stable infrastructure and dominant languages, which do not reflect fragile or conflict-affected environments where PeaceTech is often applied. Trust is another major issue. Practitioners often recognise the potential of technology but have limited confidence in how tools are designed and introduced. In PeaceTech contexts, concerns include misuse of sensitive data, misinformation, lack of contextual understanding, and the erosion of human relationships. At the same time, there are clear opportunities. There is growing recognition that AI governance must be more human-centric and context-aware. This aligns with emerging PeaceTech approaches that prioritise local participation, shared ownership, and the principle of Do No Harm. Addressing these gaps could help ensure that AI supports resilience and peace rather than reinforcing inequality or risk.

The AI Dialogue can act as a bridge between global governance discussions and practical realities. It can connect high-level policy frameworks with field-based experience, ensuring governance reflects challenges such as data access, linguistic diversity, trust, and safety in fragile environments, including those where PeaceTech applications are emerging. It can also support international cooperation by identifying shared principles while recognising that implementation must remain context-specific. This is particularly important for PeaceTech, where tools must adapt to highly varied social and cultural conditions. Importantly, the Dialogue can bring together actors who do not usually engage equally. Practitioners, civil society, and local actors often have direct knowledge of risks but remain underrepresented, despite being central to PeaceTech implementation. By creating structured, inclusive exchange, the Dialogue can make cooperation more grounded and actionable.

The Dialogue should build on frameworks developed by UNESCO, UN digital cooperation efforts, and human rights-based approaches to data governance. It should also draw from humanitarian and peacebuilding practices, particularly around data responsibility and conflict sensitivity, which are directly relevant to PeaceTech. In addition, emerging multi-stakeholder initiatives offer practical models. In Austria, collaborative approaches bring together government, academia, peace organisations, and technical actors to explore human-centric AI governance. This includes work through the PeaceTech Alliance. These efforts include defining PeaceTech as the use of hardware, software, and data systems across the conflict cycle to support mediation, anticipate risks, strengthen peace operations, and foster trust, grounded in Do No Harm. The added value of the Dialogue is to connect these efforts and translate principles into coordinated global practice.

Different stakeholders should contribute based on their strengths. Governments provide policy direction, technical actors provide system expertise, and civil society and practitioners contribute insights on impact and trust. This is particularly important in PeaceTech contexts, where practical experience of deployment in fragile environments is essential for informing governance. The Dialogue should combine plenary sessions with smaller working groups, practitioner panels, and evidence-based workshops. Participation should not be hierarchical. Community organisations and practitioners should be recognised as governance actors, especially where they are directly involved in PeaceTech implementation. A layered format with global, regional, and thematic discussions would support meaningful engagement.

Underrepresented voices include civil society organisations in the Global South, community-based actors, peacebuilding practitioners, humanitarian workers, and non-English-speaking communities. These groups are often closest to the real-world impacts of AI systems but remain farthest from governance processes. In many cases, these actors are not excluded due to a lack of knowledge, but because their expertise is not recognised within current governance structures. Local organisations and community actors often hold rich, context-specific knowledge, including social dynamics, conflict drivers, cultural practices, and informal communication systems, yet are rarely treated as contributors to system design or governance. One reason for this gap is that governance processes tend to prioritise technical and institutional expertise over community-based knowledge. However, in practice, effective AI systems, particularly in sensitive contexts, depend on understanding how communities organise, communicate, and build trust. Approaches drawn from community development can offer useful entry points here. For example, just as communities understand and manage shared physical assets such as schools, religious institutions, or community centres, similar thinking can be applied to data as a shared community asset, where local actors have a role in how it is collected, used, and governed. This perspective is increasingly reflected in emerging research and practice, including recent work examining community-driven approaches to AI data in sub-Saharan Africa (I have a Cambridge article on this). To include these voices meaningfully, governance processes should support multilingual engagement, remote participation, financial accessibility, and formats that do not rely solely on technical expertise. More importantly, they should recognise community actors as co-designers and governance participants, not only as beneficiaries or data sources.

To foster meaningful and dynamic engagement, the Dialogue should move beyond traditional panel formats and create spaces where participants can work through real-world governance challenges together. One effective format would be scenario-based labs, where participants are given concrete cases, for example AI use in conflict-affected environments, misinformation dynamics, or data governance in low-resource settings, and asked to collaboratively identify risks, trade-offs, and governance responses. This shifts discussion from abstract principles to applied decision-making. A second approach would be practitioner hearings, where frontline actors present short, evidence-based reflections from their work, followed by structured responses from policymakers and technical experts. This creates a direct link between lived experience and governance discussion, rather than treating practitioners as peripheral voices. Co-design sessions could also be used, bringing together mixed groups of stakeholders to map how an AI system would be introduced into a specific context, including data collection, community engagement, risk mitigation, and oversight. This would help reveal where governance assumptions break down in practice. In addition, the Dialogue should support asynchronous and remote participation, allowing contributions from those unable to attend in person, particularly from underrepresented regions and communities. Finally, live synthesis and feedback loops should be built into the process, where key insights from each session are immediately captured, shared, and refined with participants. This increases transparency and ensures contributions shape outcomes in real time. These formats would make the Dialogue more interactive, grounded, and inclusive, supporting genuine exchange rather than parallel statements.

Effective AI governance can build on a combination of established normative frameworks and emerging practice-based approaches. Human rights-based frameworks, such as those developed through UNESCO, provide a strong foundation by centring dignity, accountability, inclusion, and non-discrimination. These are essential for ensuring that AI governance addresses not only technical performance but also social impact and inequality. In parallel, practices from the humanitarian and peacebuilding sectors offer valuable operational models. Approaches such as Do No Harm and data responsibility frameworks emphasise context awareness, risk mitigation, and the protection of sensitive information, particularly in fragile or high-risk environments. These principles are increasingly relevant as AI systems are deployed in more complex social settings. Community-based data governance models also offer concrete solutions. Treating data as a shared community asset, where local actors have a role in how data is collected, interpreted, and used, can help address issues of trust, legitimacy, and perceived data extraction. This approach aligns governance with local realities and strengthens participation. In practice, multi-stakeholder platforms are beginning to operationalise these principles. In Austria, the PeaceTech Alliance brings together government, universities, peace organisations, and technical communities to collaboratively explore how AI and data can support peacebuilding. This includes work on defining PeaceTech as the use of hardware, software, and data systems across the conflict cycle to support mediation, anticipate risks, strengthen peace operations, and foster trust, grounded in the principle of Do No Harm. Such approaches demonstrate how governance can move from abstract principles to coordinated, practice-based implementation.