Knowledge Satnav

A success outcome is to set up a community to explore standards for the "rules of rules" — i.e., governing rule governance itself: how governance rules are written, interpreted, updated, overridden, and logged, and how different rule systems remain interoperable across institutions and jurisdictions. A second success outcome is to surface and align on a human–machine boundary framework that is operationalizable. The Dialogue should move beyond abstract principles and enable task-level allocation logic that is usable in real workflows: what stays human-led, what can be AI-led, under what conditions, and what oversight strength is proportional to the risk. A third success outcome is a reframing: stop focusing on managing AI mainly through digital/structured-data rules alone. We need to rediscover forgotten human institutional wisdom for governing delegated capability—contracts, laws, policy language, mandates, separation of powers, escalation rights, due process—and translate that into AI-era operational governance. The Dialogue should explicitly treat AI interoperability with human governance as a core requirement for legitimacy and trust.

• Interoperability of governance approaches
• Social, economic, ethical, cultural, linguistic and technical implications of AI
• Transparency, accountability, and human oversight
• Open-source software, open data and open AI models

Interoperability of governance approaches is urgent because the world will not converge on a single governance model. What we need is the "rules of rules" layer-governing rule governance-so different approaches can connect, remain legible to each other, and avoid fragmentation. Transparency, accountability, and human oversight are urgent because we need to operationalize values instead of tokenistical ideas. That means governance must specify decision rights and control paths: who is accountable, who can override, what gets logged, and how escalation works in practice. Open-source software, open data, and open AI models matter because they shape whether governance is real or performative. Without meaningful access, many actors cannot evaluate, adapt, audit, or locally govern systems, which worsens the AI divide and undermines legitimacy. The broader social/economic/ethical/cultural/linguistic/technical implications theme matters because governance must remain credible across cultures and contexts, not only technically correct.

Rules of rules (governance of governance): a shared way to govern how governance rules are created, revised, interpreted, overridden, and audited across different institutions and jurisdictions. Operationalizing values instead of tokenistic ideas: turning principles into actionable governance mechanics (decision rights, escalation paths, override logic, logging requirements, accountability chains). AI interoperability with human governance is necessary: ensuring AI governance plugs into existing human institutional tools (contracts, laws, policy language, mandates, separation of powers, due process), not only technical compliance layers. AI should be governed by unstructured-data rules because LLM-based AI is a new kind of capability: it is natively trained on and operates through unstructured human knowledge (language, documents, policy, contracts, culture). Even when it produces structured outputs, the competence is learned from unstructured corpora and expressed through context-heavy judgment. That means governance cannot rely only on structured data rules and digital controls; it must also be expressible in the same unstructured rule mediums humans use to govern institutions (law, policy, contractual obligations, procedural constraints), so oversight remains legible, enforceable, and interoperable across cultures and systems.

Governance gaps in the thematic areas I selected are already creating significant stress in my sector: strategic-level governance and institutional legitimacy for human collectives. The core gap is governance delay under acceleration. In the past, governance cycles and technology cycles were roughly comparable; now policy and institutional adaptation still move in long cycles while new AI capabilities emerge in much shorter cycles. This widening gap is not a temporary coordination issue — it challenges governance mechanisms built on older assumptions about the pace of change. The most significant challenges are: fragmentation and interoperability failure (different actors adopting incompatible governance approaches); weak operational accountability and oversight (principles exist, but decision rights, escalation paths, and auditability are not consistently embedded into real workflows); and legitimacy risk (public trust declines when institutions cannot explain, control, or visibly govern the effects of AI on social and economic outcomes). Open-source/open-data/open-model dynamics amplify both sides: they can reduce dependence and widen participation, but uneven access and capacity deepen the AI divide and weaken governance credibility across regions. The opportunity is that acceleration forces a fundamental rethink: governance must evolve from static, slow-moving rule-making toward adaptive "rules of rules" — governance that can update governance safely, transparently, and interoperably. If the Dialogue helps align on operational boundary logic (human vs machine), and on governance primitives that translate into law/policy/contract language, it can strengthen legitimacy while keeping pace with AI change.

The AI Dialogue can advance international cooperation by acting as an interoperability and discovery layer, not simply another venue that repeats established frameworks. Many existing processes already produce principles, standards language, and formal "best practice" outputs. The cooperation gap is that these efforts often remain fragmented, and the process incentives can over-weight familiar, well-resourced, already-visible actors—creating a risk of a consensus trap: procedurally legitimate discussion that becomes operationally irrelevant under rapid capability change. To add real cooperative value, the Dialogue should be designed to surface mechanism-level governance insights (how accountability, oversight, escalation, and auditability work in practice) and translate them into shared, portable building blocks that different jurisdictions can adopt without forcing convergence on a single governance ideology. It should also strengthen legitimacy by making governance visible and credible to publics, not just compliant on paper. Practically, the Dialogue can: (1) establish transparent criteria for what counts as "high-leverage" governance contributions (operational primitives, not slogans), (2) deliberately elevate underrepresented but high-novelty work alongside established initiatives, and (3) connect promising ideas into existing channels (standards bodies, regional processes, capacity-building partnerships) without diluting them into generic language. In short: cooperation improves when the UN Dialogue reduces fragmentation, prevents capture by incumbency, and builds shared governance interfaces that keep pace with acceleration.

The AI Dialogue should build upon and connect with existing initiatives that already shape global AI governance, including: international and national standards and management-system efforts; risk and assurance frameworks used by governments and industry; intergovernmental principles and regional processes; human-rights and ethics mechanisms within the UN system; and technical safety, evaluation, and incident-learning communities. These initiatives collectively provide important normative anchors and technical workstreams, but they often operate in parallel, with limited interoperability and unequal visibility for under-resourced regions and novel approaches. The added value the AI Dialogue can bring is not to compete with these initiatives, but to provide the missing connective tissue: (1) interoperability across governance approaches—shared language and shared evidence artifacts for transparency, accountability, and human oversight; (2) a discovery function that prevents "procedural legitimacy" from substituting for operational progress by ensuring the Dialogue does not become only a showcase of already-dominant frameworks; and (3) a structured pathway to route high-value, underrepresented contributions into the right existing mechanisms (standards, policy coordination, capacity-building) so they can be tested, adapted, and scaled. Concretely, the Dialogue could publish transparent agenda and summary criteria, reserve space for underrepresented high-novelty mechanism design, and allow minority insights to be recorded so useful divergence is not erased by consensus. This would strengthen legitimacy, reduce fragmentation, and make international cooperation more practical and inclusive.

Different stakeholders should contribute, but the Dialogue must avoid "fake fairness" where inclusion means equal airtime and the most legible, well-resourced voices still dominate. Inclusivity should mean equal access to being evaluated on merit, and the format should be designed to surface underrepresented, high-insight contributions that are otherwise filtered out by institutional packaging. Recommendations for format and structure: • Use a merit-based intake that scores inputs on insight density, operational usefulness, interoperability value, and novelty—not prestige or how polished the submission is. • After selection, provide support for legibility (structured templates, editorial/translation assistance, coaching) so high-value work is not disadvantaged for being hard to explain quickly. • Organize breakouts around mechanisms and failure modes (what actually works, what breaks, what governance primitives transfer across contexts), not general statements. • Make outcomes traceable: the Co-Chairs' summary should clearly reflect which inputs were elevated and why, so the process rewards merit and avoids capture. This structure allows stakeholders to contribute based on what they uniquely know, while ensuring the Dialogue actually discovers and elevates high-leverage governance ideas.

Underrepresented voices are often not simply "missing," but systematically filtered out because global governance discussions reward legibility, institutional branding, and established networks. The most underrepresented perspectives include: contributors from low-capacity contexts; local-language and culturally specific communities; frontline operators (procurement, regulators implementing policy, auditors/assurance); impacted workers and public service delivery contexts; and independent technical/public-interest communities outside major labs. The most important underrepresentation is also a type of contribution: high-novelty, high-insight work that is not yet widely known and is harder to communicate in standard policy language. There is also a systematic cultural bias where governance solutions from non-dominant language communities are dismissed as "we already tried that." More accurately, English-dominant governance communities tried a version of it inside their own institutional assumptions. Communities operating deeply within other legal-cultural traditions may hold practical governance mechanisms that the dominant discourse cannot easily imagine, and these are often filtered out by arrogance, legibility bias, or prestige networks. Inclusion should therefore be merit-corrective, not "equal allocation." Practical ways to include underrepresented high-value perspectives: create a protected high-insight/high-novelty pathway with transparent criteria; provide multilingual and participation enablement plus editorial support; and ensure outcomes record "minority insights" so consensus does not erase valuable divergence.

Innovative engagement formats should be designed to prevent the AI Dialogue becoming a stage for "pitching up" to the UN. The UN's comparative advantage is to enable a high volume of multi-stakeholder interaction, then use AI-assisted sensemaking to identify patterns, convergence, and underrepresented high-insight ideas that are usually filtered out by legibility bias or prestige dynamics. Formats that would foster meaningful and dynamic engagement: • Parallel micro-dialogues (many small groups): short, facilitated discussions across the four themes with rapid rotation, so participation is not dominated by a few voices and ideas can recombine across sectors and regions. • Interlinking sessions (from speeches to connections): participants must link their point to others in real time (supports / contradicts / extends), producing a visible map of relationships rather than isolated statements. • Mechanism clinics (focus on what actually governs): sessions where contributors present one concrete governance mechanism or practice (e.g., oversight decision rights, escalation paths, audit artifacts, cross-border accountability handling) and others stress-test it across contexts for interoperability. • AI-assisted synthesis wall: AI clusters inputs and discussion outputs live into themes/mechanisms/failure modes; participants validate and refine the clusters; the Dialogue captures "what emerged" rather than "who spoke." • Protected "high-novelty" channel: a structured way to retain unfamiliar but potentially high-leverage ideas so they are not lost to time constraints or premature consensus. These formats keep the Dialogue inclusive without becoming performative: they increase interaction, preserve traceable outputs, and help the UN extract cooperation-relevant building blocks from the full field of discussion.

One set of approaches I recommend can be understood as a stack, with an overarching standard and four interoperable governance building blocks beneath it: 1) Open Governance Standard (OGS) - overarching A unifying frame for "governance as a system," intended to make governance approaches comparable and interoperable across institutions and jurisdictions. Its value is providing a top-level structure for how governance components relate, so cooperation doesn't collapse into fragmented checklists. https://github.com/kfkchau/Open-Governance-Standard/blob/main/pitch-deck.pdf 2) Open Meta-Governance Standard (OMGS) - rules of rules - Governs rule change A meta-governance approach: governing rule-governance itself (how rules are written, interpreted, updated, overridden, and logged). This helps address the real gap in most AI governance work: governance often exists, but governance-of-governance is unclear, making interoperability and legitimacy fragile. https://github.com/kfkchau/Open-Meta-Governance-Standard 3) Common Governance Language (CGL) - shared grammar - described governed reality A shared vocabulary/grammar for governance concepts so different regimes can "translate" between each other. This is practical for international cooperation because it reduces ambiguity and improves cross-border legibility without forcing a single governance ideology. https://github.com/kfkchau/Common-Governance-Language 4) Task Boundary Framework (TBF) - operational human-machine boundary - defines governable work A task-level method to allocate work between humans and AI using reasoning tiering and data structurality, enabling proportional oversight and auditability. This converts abstract "human oversight" into actionable allocation logic. https://github.com/kfkchau/project-ahi-fin/blob/main/subfw-ahibom/pitch-deck.pdf 5) Project White Collar Framework (PWC) - governance embedded in operations - execute that work in a governed way An operational architecture that embeds accountability into execution via diversity of perspective, separation of powers, collective decision making, and immutable logging primitives-turning governance principles into enforceable workflow mechanics. https://github.com/kfkchau/Project-White-Collar/blob/main/pitch-deck.pdf Together, this stack operationalise interoperability, accountability, and legitimacy while remaining adaptable across sectors and capacity levels.