ICT4Peace

Consolidation of ethical standards and outputs that support practical implementation and usage of those standards.

• Safe, secure and trustworthy AI
• Interoperability of governance approaches
• Transparency, accountability, and human oversight
• Protection and promotion of human rights

It's difficult to choose, but first and foremost we need to focus on safety and the means to effectively implement and oversee it. This is why having more joined up approaches to governance and meaningful human oversight are necessary in order to have trustworthy AI that protects human rights.

While human oversight is mentioned, what is key is that it is meaningful human oversight--not just a human that checks the boxes or clicks "proceed," but who actually reviews AI recommendations and is able to change or abort them when necessary.

Many governance gaps arise from the fact that existing legal and regulatory frameworks have not caught up to the advances in technology. Technologies are deployed because they are technically feasible and there is market demand, not because they are in line with, for example, human rights' standards. Translation of existing obligations, laws and regulations is of utmost importance in order to safeguard the regulatory frameworks we already have.

Facilitate "translation" of international obligations with regard to AI, and support development of interoperable governance standards that apply across borders.

The Geneva AI Summit in 2027 is a key event, with many workshops and other work streams taking place in the coming months before the Summit.

Governments are regulators, and can develop and implement regulations that effectively govern use of AI. Civil Society can provide a "ground-level" view of the lived impacts of AI, and also provide feedback as to whether regulations are actually effective. Companies have "effective control" over the development of technologies, and have the power to change elements that are having negative impacts. Standards bodies can help with interoperability of legal, ethical and technical standards. The AI Dialogue should be structured to make these contributions usable. I recommend a format with: 1. a high-level plenary for political direction; 2. thematic working groups focused on safety, human rights, interoperability, capacity-building and meaningful human oversight; 3. expert roundtables for technical and legal deep dives; 4. multi-stakeholder consultations with public comment opportunities and meaningful ability to impact decision-making; and 5. a lightweight synthesis mechanism that turns inputs into concrete outputs, not just summaries. To avoid dominance by a few actors, speaking time, written input formats and chairing roles should be balanced across stakeholder groups and regions. Sessions should be hybrid, multilingual, accessible, and designed to support both global participation and focused expert work.

Several perspectives remain underrepresented in global AI governance discussions. These include: communities in the Global South; least developed countries; small island developing states; Indigenous peoples; local and municipal authorities; persons with disabilities; older persons; children and youth; women and gender-diverse communities; workers and trade unions; frontline public servants; humanitarian and human rights practitioners; and communities most directly affected by AI in migration, policing, welfare, education and employment contexts. The Dialogue should support plain-language submissions, multilingual participation, and remote access options that do not assume high bandwidth or specialist knowledge. Community-based organizations and local experts should be supported to convene preparatory consultations before global sessions. Finally, the Dialogue should actively seek inputs from those affected by AI harms, not only from those designing or regulating systems. Inclusion should be treated as a governance issue, not an outreach afterthought. Without broader participation, global AI governance risks reproducing existing inequalities rather than correcting them.

Engagement formats that encourage and enable practical problem solving and public participation, such as: - Scenario-based workshops on real-world AI use cases, such as public-sector decision-making, frontier model safety, health, education or humanitarian response. - Structured "crosswalk labs" where participants compare existing standards and identify convergence, gaps and conflicts. - Policy simulation exercises that let stakeholders test how different governance approaches perform under pressure. - World-café or rotating small-group formats to enable more equal participation and reduce dominance by a few voices. - Expert clinics where states, civil society or smaller organizations can receive practical advice on implementation. -Demo-and-critique sessions for tools, audits, assessments or governance platforms. - Youth and community assemblies linked to the formal Dialogue, with direct routes for their inputs to influence outcomes. - Public-facing digital consultation spaces with moderated discussion, summaries and traceable feedback loops. A particularly useful format would be a "problem-to-practice" session: one stakeholder presents a concrete governance challenge, and mixed groups develop a short action-oriented response. The Dialogue should also publish back what it heard and how it used the input, so participants can see impact. Engagement should be iterative, multilingual, hybrid, and accessible to those with limited resources. The goal should be not only inclusion, but genuine deliberation and practical outcome generation.

Examples of policies, practices, platforms, and approaches that, and approaches that promote effective AI governance already exist, but they are fragmented and unevenly applied. At the policy level, the EU AI Act offers a risk-based model that distinguishes between prohibited, high-risk, and lower-risk uses, with specific obligations for transparency, oversight, and accountability. The Council of Europe Framework Convention on AI provides an important human rights-, democracy-, and rule of law-based approach. The UNESCO Recommendation on the Ethics of AI and OECD AI Principles offer globally relevant normative baselines, while the UN Global Digital Compact reinforces the need for an open, safe, inclusive, and human-centred digital future. At the implementation level, the NIST AI Risk Management Framework provides a practical lifecycle approach to identifying, measuring, and managing AI risks. ISO/IEC 42001 and ISO/IEC 23894 offer management-system and risk-management standards that can support organizational accountability. The IEEE 7000-series shows how ethics can be embedded into system design. Sectoral guidance, such as the WHO's work on AI in health, demonstrates the value of context-specific governance. A particularly promising approach is to create accessible standards maps and governance toolkits that translate complex frameworks into practical guidance for different actors. This would help bridge the gap between principles and implementation, reduce duplication, and support more consistent, rights-respecting AI governance across sectors and regions.