UNGP for SDs

For the first Global Dialogue on AI Governance to be a success, it must move beyond high‑level principles to deliver three measurable, accountability‑anchored outcomes that integrate human rights, mental health, and cross‑sector governance. 1. Human Rights and Mental Health as Core Safeguards AI governance must operationalize the right to mental health. Psychological and social harms—such as algorithmic manipulation, cognitive erosion, and AI‑driven workforce displacement—are not secondary effects; they are fundamental human rights issues. A clear success would be the adoption of a framework that mandates psychological impact assessments for high‑risk AI systems, embedding mental well‑being into the risk‑management cycle alongside privacy and non‑discrimination. 2. Accountability and Financing as a Governance Mechanism Drawing on over a decade of designing cascading accountability models for SDG implementation, I have learned that pledges without enforceable follow‑up rarely translate into impact. A key win for the Dialogue would be to launch a multi‑stakeholder accountability mechanism—backed by a voluntary SDG‑aligned finance instrument—that ties access to compute, data, and innovation capacity to verifiable human rights and environmental due diligence. 3. From Ad‑Hoc Reporting to a Binding Accountability Framework Success ultimately means moving from voluntary, ad‑hoc reporting to a structured, binding accountability framework that includes independent third‑party verification. This is where the UNGP Endorsed Partner Programme's cascading model can serve as a practical blueprint for embedding responsible AI requirements throughout the AI lifecycle—from design and deployment to procurement and public administration.

• 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

As a UN Special Rapporteur candidate and a psychologist, I prioritize human rights because AI systems can directly impact mental health, dignity, and non-discrimination-areas often overlooked in technical debates. Embedding human rights from the design stage is non-negotiable. Transparency, accountability, and human oversight form the operational backbone. Drawing on my experience with the UNGP cascading accountability model, I know that voluntary pledges fail without enforceable mechanisms. Success requires binding, verifiable standards with independent oversight. AI capacity-building is essential for global equity. Developing nations must not be passive recipients of AI; they need the skills, infrastructure, and policy frameworks to govern AI for their own social and economic priorities. This is SDG 17 in action. Finally, the social, economic, ethical, cultural, linguistic and technical implications cannot be treated as separate silos. AI governance must be interdisciplinary and context-aware. A purely technical approach misses the behavioural, economic and cultural dimensions-domains where psychology, finance and international law intersect. Together, these four pillars ensure that AI serves humanity, not the reverse.

While the listed themes cover essential ground, two cross-cutting issues deserve explicit attention. Mental health as a distinct pillar - Although human rights protection is included, the psychological dimension of AI - algorithmic manipulation, emotional profiling, AI-driven social comparison, and cognitive autonomy - is rarely addressed in governance frameworks. As a psychologist and UN Special Rapporteur candidate, I emphasise that mental well-being is not a subset of privacy or non-discrimination; it requires its own safeguards, including mandatory psychological impact assessments for high-risk AI systems. Financing and economic incentives for accountable AI - Transparency and oversight are vital, but without aligning financial flows, they remain weak. Drawing on my finance background, I see a gap: there is no mechanism linking AI governance compliance to access to capital, compute, or data. An emerging issue is the need for SDG-aligned, multi-stakeholder financing instruments that reward responsible AI development and penalise systemic non-compliance - a practical application of SDG 17 (Partnerships for the Goals). This would transform accountability from a regulatory burden into a strategic advantage.

In Canada, the governance gap is not a lack of AI principles but a lack of enforceable, cross‑sectoral accountability. The rapid adoption of AI in healthcare, finance, and public services has outpaced binding oversight. As a result, vulnerable communities—including youth and those with mental health conditions—face risks of algorithmic discrimination without effective remedy. This directly affects my work in SDG accountability: without a cascading model that ties AI due diligence through supply chains, voluntary human rights pledges remain toothless. Globally, the divide between AI‑capable and AI‑dependent nations widens capacity gaps. As a UN Special Rapporteur candidate, I see that without structured capacity‑building (SDG 17), developing countries cannot shape AI governance on their own terms. The opportunity lies in creating interoperable, verifiable standards that bridge North‑South asymmetries and embed psychological impact assessments as a core requirement. On the positive side, advances in transparency tools (model cards, audit trails) offer a foundation. The opportunity is to turn them into legally anchored, independently verified requirements. My sector—SDG accountability—is uniquely positioned to pilot such frameworks, leveraging the UNGP Endorsed Partner Programme's cascading model to transform AI governance from a patchwork of guidelines into a measurable, human‑rights‑centred system. This would turn today's governance gap into tomorrow's competitive advantage for responsible AI innovation.

The first Global Dialogue on AI Governance can serve as a catalyst for interoperable, accountable cooperation by shifting the focus from shared principles to shared enforcement mechanisms. Drawing on my work with the UNGP cascading accountability model, I see three concrete roles. First, establish a common framework for binding due diligence. The Dialogue should move beyond voluntary codes of conduct and agree on a minimum set of verifiable standards covering human rights, mental health, and environmental impact. This would create a baseline that all member states can adopt, preventing a race to the bottom. Second, create a multi‑stakeholder verification mechanism. As a finance professional, I know that trust requires independent audit. The Dialogue can mandate a third‑party assurance process for high‑risk AI systems, modelled on existing ESG reporting standards. This would provide credibility and reduce fragmentation. Third, launch a capacity‑building alliance for the Global South. International cooperation is hollow without equity. The Dialogue should establish a dedicated fund and technical assistance programme to help developing nations draft AI legislation, train regulators, and build safe AI infrastructure. This directly supports SDG 17 (Partnerships for the Goals). Ultimately, the Dialogue's unique value is to transform AI governance from a collection of isolated national efforts into a coherent, accountable international regime – one where human rights, mental well‑being, and enforceable transparency are not afterthoughts, but the architecture itself.

The first Global Dialogue on AI Governance should build upon and interconnect several existing initiatives, transforming them from a fragmented collection of principles into a coherent, accountable international regime. Key initiatives to build upon First, the OECD AI Principles provide a mature, values‑based framework (human‑centred values, transparency, robustness, accountability and inclusive growth) with five corresponding policy pillars. The EU AI Act, which is already being implemented, offers the world's first comprehensive, risk‑based regulatory blueprint, including binding rules for high‑risk systems and a governance architecture with national competent authorities and an EU‑level AI Office. The UNESCO Recommendation on the Ethics of AI supplies globally agreed ethical guardrails and policy action areas that directly address mental privacy and human dignity. The UN Guiding Principles on Business and Human Rights (UNGPs), which are increasingly being applied to the AI lifecycle, establish a baseline for corporate human rights due diligence, access to remedy and State obligations. What the AI Dialogue can add The unique value of the Dialogue is to provide the missing accountability and interoperability layer. It can: · Transform voluntary principles into enforceable, verifiable standards by adopting binding due diligence and independent third‑party assurance, building on the proven cascading accountability model. This would give operational teeth to the existing frameworks. · Bridge the gap between human rights law and AI practice by mandating psychological impact assessments, moving mental well‑being from an afterthought to a core governance requirement. · Establish a multi‑stakeholder verification body responsible for accrediting national and corporate compliance schemes, thereby ending the current fragmentation of ad‑hoc reporting. Crucially, the Dialogue has the authority to report directly to the General Assembly, giving it a political weight that other initiatives lack. Its principal added value is to convert today's patchwork of soft principles into a hard, harmonised accountability architecture where human rights, mental health, and enforceable oversight are not optional extras but the very foundation of international AI governance.

To be effective, the AI Dialogue must move beyond traditional plenary speeches and adopt a multi‑stakeholder, action‑oriented format that mirrors the complexity of AI governance itself. Recommended structure I propose three interconnected tracks running in parallel: 1. Government Track – A negotiating table for binding commitments, focusing on interoperability of standards, mutual recognition of conformity assessments, and financing for capacity‑building in the Global South. Closed‑door sessions would allow candid exchange on red lines and trade‑offs. 2. Expert Track – Thematic working groups on human rights (including mandatory psychological impact assessments), transparency and independent audit, and cross‑border data flows. These groups would produce draft model provisions and voluntary "playbooks" for national regulators. 3. Stakeholder Forum – Open sessions for civil society, youth networks, indigenous representatives, and business. This forum would feed lived experience and practical constraints into the negotiating track, ensuring that the final outcomes are grounded in reality, not abstraction. How stakeholders can contribute · Governments must commit to specific, time‑bound deliverables (e.g., a baseline due diligence law by 2028) and to sharing enforcement data. · Private sector should demonstrate existing accountability mechanisms (e.g., third‑party audits, algorithmic impact assessments) and commit to public registries of high‑risk AI systems. · Civil society and academia should provide independent monitoring, mental health expertise, and grassroots evidence of harm. · International organisations (UNESCO, OECD, ITU, UNGP) should offer technical secretariats, model laws, and capacity‑building programmes. To avoid a talking shop, each track should produce a concrete output (a template law, a verification protocol, a funding pledge) and agree on a follow‑up mechanism – for example, annual progress reviews under UN General Assembly auspices, with a mandate to escalate persistent non‑compliance. Only then will the Dialogue be worthy of its name.

Several critical perspectives remain systematically underrepresented in global AI governance discussions. Underrepresented voices First, mental health professionals and psychologists. While privacy and non‑discrimination receive attention, the psychological impacts of AI – algorithmic manipulation, cognitive erosion, social comparison, and AI‑induced anxiety – are rarely addressed. As a Chartered Psychologist and UN Special Rapporteur candidate, I see this gap as both dangerous and easily remediable. Second, Global South communities, particularly grassroots civil society organisations and indigenous groups. Most high‑level forums are held in Geneva, New York or Brussels, with English‑only proceedings and technical jargon that exclude local advocates. Their lived experiences of algorithmic bias, data extraction and labour exploitation are essential for just governance. Third, youth and intergenerational voices. AI will shape the world young people inherit, yet their participation is often tokenistic. Meaningful inclusion requires decision‑making power, not just speaking slots. Fourth, non‑English speakers and culturally diverse epistemologies. AI governance frameworks currently reflect Western legal and ethical traditions. Indigenous knowledge systems and non‑Western concepts of relationality and collective well‑being are largely absent. How to include them The AI Dialogue can remedy this through concrete design choices: · Dedicated funding for travel, interpretation and remote participation for Global South representatives. · Mandatory inclusion of mental health experts on each thematic panel, with a requirement to report on psychological impact assessments. · A Youth Council with co‑decision authority, not merely advisory, modelled on successful UN youth engagement mechanisms. · Multilingual proceedings and culturally adapted materials, with interpretation into at least all six UN official languages and, where feasible, major regional languages. Inclusion is not charity; it is a prerequisite for legitimate and effective AI governance. Without these voices, we risk building a global regime that works for the few, not the many.

To foster meaningful and dynamic engagement, the AI Dialogue must abandon passive panel‑centric formats and embrace interactive problem‑solving. Three innovative formats stand out. 1. Accountability Labs (reverse hackathons) Participants would be assigned a real‑world AI governance failure (e.g., an algorithmic discrimination case) and given a limited time to design a verifiable accountability mechanism – including monitoring, audit and remedy. Teams would present their solutions to a jury of peers, with the best proposals published as model protocols. This format leverages participants' collective expertise and shifts the focus from talking to building. 2. Human Rights Impact Simulation Drawing on my psychology background, this immersive activity would place decision‑makers in the shoes of individuals affected by high‑risk AI systems – a job applicant rejected by an opaque algorithm, a patient misdiagnosed by an AI triage tool, or a young person experiencing AI‑driven mental distress. After the simulation, each table must produce a concrete "human rights safeguard" for a specific stage of the AI lifecycle. This transforms abstract principles into actionable empathy. 3. Thematic Sprint Sessions Instead of open‑ended discussions, each working group would be tasked with drafting a single, specific deliverable in a 90‑minute sprint: a template law, a verification checklist, or a funding pledge mechanism. Drafts would be reviewed and iterated by another group, creating accountability and cross‑pollination. 4. Youth‑Led Pre‑Dialogue Assembly A full day before the main event, the Youth Council (with co‑decision authority) would generate recommendations and submit them directly to the government track. Their proposals would be marked as "youth‑originated" in final outputs, ensuring visibility and ownership. These formats convert passive attendance into active co‑creation, leaving participants not with a vague communiqué, but with a shared sense of agency and a set of testable, accountable commitments.

Several concrete examples demonstrate effective AI governance moving from principles to accountability. Human rights-centred frameworks - The Council of Europe's HUDERIA methodology enables organisations to conduct structured, rights-based AI risk assessments, moving beyond cybersecurity to human dignity. UNESCO's AI Recommendation, adopted by 194 states, bans social scoring and mass surveillance and mandates ethical impact assessments. Binding accountability mechanisms - The EU AI Act provides a risk-calibrated, enforceable blueprint with fines, third-party conformity assessments, and national oversight authorities. The Council of Europe's Framework Convention on AI (March 2026) is the first legally binding international AI treaty, obligating states to ensure public and private AI systems respect human rights, democracy and the rule of law. Independent audit and cascading due diligence - Platforms such as Resaro and SeekrGuard offer third-party AI auditing and certification. Drawing on UNGP's cascading accountability model, the same mechanism can require foundation model providers to extend human rights and psychological safety obligations through their entire AI supply chain. Multi-stakeholder and Global South leadership - India's "Samaj, Sarkar, Bazaar" triadic model co-designs AI rules with society, government and markets. KICTANet elevates collaborative governance from the Global South, ensuring developing nations are co-authors, not passive recipients of norms. The EU's AI sandboxes allow stress-testing of regulations before they become binding. Together, these examples show that effective governance is achievable. The missing element is the political will to embed rights-centred impact assessments, mandatory third-party audit, cascading supply chain due diligence and binding treaty commitments into a coherent global system.