Open Government Partnership

Success would mean the Dialogue moves beyond high-level principles and produces a shared, practical agenda that helps countries govern artificial intelligence in the public interest. From an open government perspective, that includes: (1) clearer agreement on minimum expectations for transparency (e.g., public information on where AI is used in government, for what purpose, and who is accountable); (2) practical pathways for accountability and oversight (impact assessments, auditability, complaint/redress channels, and independent review); (3) commitments to inclusive participation—ensuring civil society and affected communities can shape AI policy, procurement, and deployment decisions, not only technical experts; and (4) a stronger, resourced focus on implementation support and capacity-building so that administrations can operationalize safeguards without being locked out of AI's potential benefits. A successful Dialogue would also identify a small set of implementable "building blocks" that countries and institutions can adopt immediately (e.g., AI/algorithm registers, accountable procurement checklists, participatory impact assessment approaches). This is especially important given the growing pace of government AI adoption and the risk that governance arrives only after harms occur. Outcomes should reflect human rights obligations, prioritize equity and inclusion, and emphasize public trust as a prerequisite for sustainable AI adoption.

• 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

The Open Government Partnership's (OGP) experience suggests that the central challenge in AI governance is not a lack of principles, but an "implementation gap": governments are adopting AI faster than effective safeguards are being put in place. OGP's approach therefore prioritises transparency, accountability, and human oversight because these are the practical mechanisms that make AI governance real-through public visibility of AI use, clear responsibility, independent scrutiny, and routes for contestability and redress. We prioritise human rights because government AI systems can affect access to essential services and rights (due process, non-discrimination, privacy). Human-rights-centred governance, supported by impact assessments and ongoing monitoring, is essential to prevent harms and maintain public trust. We prioritise safe, secure and trustworthy AI because AI systems must be robust against misuse, cyber threats, and harmful failures, especially where they affect essential services. This includes practical risk management (testing and monitoring), security-by-design, clear standards for reliability and auditability, and safeguards to prevent fraud, manipulation, or unintended impacts. A focus on safety and security also helps build public trust and ensures that adoption is sustainable and aligned with democratic values and human rights. Finally, we prioritise wider social/economic/ethical implications because AI is reshaping labour markets, public services, information ecosystems, and power relationships. Governance must address real-world impacts (including bias, exclusion, and public trust) and not only technical performance. Open government approaches help ensure those impacts are debated publicly and governed in the public interest.

Several cross-cutting issues deserve stronger attention: - Implementation and learning infrastructure: Global frameworks should be paired with mechanisms for peer learning, technical support, and evidence-sharing on "what works" in real deployments. - Inclusion in agenda-setting: Underrepresentation of Global South institutions, civil society organisations, local governments, and frontline public service communities in AI governance discussions risks producing frameworks that are hard to implement or misaligned with context. - Public procurement and vendor governance: Many governments access AI through procurement. Transparent, accountable procurement-including disclosure of intended use, risk assessments, and oversight arrangements-is a critical governance lever even when formal regulation lags. - Power concentration and policy capture: The speed and scale of private investment in AI and the lobbying capacity of major firms can skew policy choices. Governance needs safeguards against capture and stronger civic oversight. - Redress and contestability in practice: Beyond "human oversight" in principle, citizens need workable mechanisms to challenge automated decisions, correct errors, and obtain remedy-particularly in high-impact public services. - Environmental and infrastructure impacts: Energy and compute costs of AI systems-especially as governments explore GenAI-should be integrated into public-interest assessments and procurement decisions.

In many contexts, government use of AI and automated decision-making is expanding in areas like eligibility determinations, fraud detection, and service targeting—often without commensurate safeguards. Where transparency is weak, systems become "black boxes" that citizens cannot scrutinise, undermining trust and increasing the risk of rights harms. Governance gaps can: - Deepen inequality through biased or poorly designed models that disproportionately affect marginalised communities; - Erode due process when individuals cannot understand, challenge, or correct automated decisions; - Increase vendor lock-in and reduce public leverage when procurement and system performance are not openly governed; - Reduce legitimacy of digital transformation as citizens perceive AI adoption as opaque or extractive. At the same time, there are opportunities: AI can strengthen access to information, improve administrative efficiency, and support integrity efforts—if governed openly. Open government approaches (transparency, participation, accountability) make it more likely that AI's benefits are realised while preventing harmful deployment and backlash that can stall innovation.

The Dialogue can function as a bridge between global discussions, emerging norms, and country-level implementation, including by aligning with existing country mechanisms and commitments on AI governance in platforms like the Open Government Partnership (OGP) and its country-level multi-stakeholder forums. This could include: - Aligning on practical governance "building blocks" (registers, impact assessments, procurement disclosure, oversight/redress) that can be adapted locally; - Creating an implementation-focused cooperation track that matches countries with peer mentors, technical support, and shared templates; - Supporting interoperability of governance approaches by encouraging comparable transparency and risk management practices, while respecting local context; - Elevating multi-stakeholder cooperation—ensuring civil society participation is treated as a core element of cooperation, not an optional add-on. This is especially valuable because many existing initiatives produce high-level principles, but fewer provide structured support and coordination across platforms to translate those principles into institutional mechanisms and day-to-day administrative practice. The AI Dialogue could also identify a small number of shared questions to be addressed via dialogue and deliberation with citizens between Dialogue sessions. This would help to ensure that global discussions are rooted in the views and concerns of the public.

Among other initiatives, the AI Dialogue can build upon the Open Government Partnership's (OGP) model of dialogue and country-owned action. As a multi-stakeholder platform with more than 70 countries and 150 local members, OGP helps countries translate high-level principles into concrete, country-owned governance reforms. Through its action plans and multi-stakeholder forums, OGP supports governments in implementing measures such as AI transparency registers, participatory oversight mechanisms, and accountability frameworks that make public-sector AI use visible and contestable. Key elements of OGP's work that the AI Dialogue can leverage include: - Country action through OGP action plans: Supporting countries to co-create and implement specific commitments such as building AI/algorithm transparency registers, institutionalising participatory governance mechanisms, and defining oversight approaches that make government AI use visible and contestable. - Bridging the "implementation gap": Recognising that global AI norms often fail at the point of delivery, OGP prioritises the recommendation of practical institutional mechanisms (e.g., disclosure requirements, impact assessments, procurement standards, oversight and redress channels) that operationalise AI governance and safeguards. - Multi-stakeholder collaboration as a core governance tool: Using OGP's model to bring government, civil society, and technical experts together so that AI policy and deployment decisions reflect public interest and human rights, as well as technical expertise. - Peer learning and global field-building: Convening practitioner exchange and producing reusable resources (templates, checklists, model approaches) so governments can adopt "building blocks" quickly and learn from each other's successes and failures. This includes supporting OGP's practitioner network of officials implementing algorithmic transparency and accountability reforms through Open Algorithms Network activities. OGP's research and experience shows that commitments in international fora work better when follow up and accountability for implementation is not an afterthought but is baked into the design of summits. Embedding the results of the AI Dialogue in OGP commitments ensures that they benefit from the multi-stakeholder dialogue, peer exchange, independent monitoring, and support that is built into OGP and which has proven to deliver results.

We recommend using a range of formats, each serving a clear purpose: - Plenary sessions for political alignment on principles and commitments - Implementation clinics focused on governance reforms and practical tools - Country-to-country peer exchanges to share experience and lessons learned - Structured civil society participation with speaking roles and co-design of outputs, not consultation alone The AI Dialogue can also identify a small set of priority questions for between-session engagement to inform future Dialogue agendas and outputs. This could include: - Global citizen dialogues (online and in-person), including citizens' assemblies to surface public priorities, risk perceptions, and "red lines" for AI governance. - National and regional multi-stakeholder dialogues with civil society organisations, industry, affected communities, local governments, and oversight institutions. - Sector-specific consultations (e.g., social protection, health, migration, justice) focused on high-impact use cases and practical safeguards. - Targeted listening with underrepresented groups (e.g. disability rights advocates, labour organisations, minority language communities, Indigenous groups, Global South practitioners). - A "what we heard / what we'll do" feedback loop where inputs are summarised publicly and linked to concrete agenda items, draft outputs, or implementation workstreams for the next Dialogue.

Underrepresented perspectives often include: communities most affected by public-sector automated decisions; disability rights organisations; labour and social protection advocates; minority language communities; Global South governments and institutions; local governments implementing AI in frontline services; civil society organisations; and practitioners working on oversight/redress in courts, ombuds institutions, and audit bodies. Inclusion measures should include: funded participation (travel and time support), multilingual materials, structured community listening sessions tied to concrete policy questions and 'what we heard' responses, and formal roles for civil society in drafting outputs. The Dialogue should also create pathways for local government and service-delivery agencies to share lessons, since many high-impact deployments happen at subnational level.

#NAME?

The Open Government Partnership's multi-stakeholder platform helps turn AI governance principles into concrete, country-owned reforms that can be implemented and tracked. Through OGP action plans and the Open Gov Challenge, governments and civil society co-create specific reform commitments with clear milestones-such as algorithm/AI registers, impact and human-rights assessments, transparent procurement practices, and independent oversight and redress mechanisms. OGP's participation standards, ongoing Support Unit guidance, and the Independent Reporting Mechanism's independent reviews help hard-wire transparency, civic participation, and accountability into how AI governance reforms are designed, delivered, and monitored over time. Examples of approaches and policies being advanced by OGP members to govern AI more openly and accountably include: Frameworks & Guidelines: - A framework to assess AI tools for bias, and the creation of complaint and redress mechanisms (Austin, Texas) - Developing a Code of Ethics on AI in public services through multi-stakeholder consultations (Dominican Republic) - Legislative framework for automated decision-making, alongside guidance for government use of generative AI (Australia) - Co-created guidance on ensuring transparency in ADM and training staff on ethical, transparent public-sector AI use (Buenos Aires, Argentina) Participatory Approaches: - Engaging human rights organisations in working groups on AI governance (Bulgaria) - Working with civil society to raise awareness of AI risks and co-create public sector AI use. - Establishing a National Multi-Stakeholder Coordination Mechanism for digital transformation and strengthening national frameworks on digital governance (Nigeria) - Developing multi-stakeholder dialogue mechanisms to design an AI governance model (Colombia) Mechanisms to increase transparency & accountability - Requiring government agencies using ADM to conduct and publish impact assessments before deployment and after systems changes are made (Canada) - Creating public algorithm transparency registers (Valencian Community, Spain; Scotland, UK); - Establishing a public oversight body for government AI use (Uruguay)