Jensen Consulting ¦ Digital Sustainability and AI

A proper recognition of the need to address how AI can keep humanity within planetary boundaries and drive forward a green circular economy.

• Social, economic, ethical, cultural, linguistic and technical implications of AI
• Interoperability of governance approaches
• Open-source software, open data and open AI models
• Protection and promotion of human rights

Social, economic, ethical, cultural, linguistic and technical implications of AI - AI is reshaping environmental governance, development planning, and scientific assessment at a pace that outstrips existing institutional capacity. From my vantage point working on digital transformation and environment, AI's social and economic implications are inseparable from its environmental ones: algorithmic decision-making in resource allocation, climate modelling, and biodiversity monitoring raises questions of equity, access, and accountability that cut across all three dimensions of sustainable development. Interoperability of governance approaches - The multilateral system has decades of experience governing transboundary, exponential challenges - from ozone depletion to climate change to plastics. These environmental treaty frameworks offer tested governance architectures (common but differentiated responsibilities, science-policy panels, compliance mechanisms) that are directly relevant to AI. Interoperability is not just a technical challenge between AI governance frameworks; it requires bridging AI governance with existing environmental, trade, and human rights regimes to avoid fragmentation and regulatory arbitrage. Open-source software, open data and open AI models - Open approaches are foundational to closing the capacity gap between countries. Environmental monitoring, early warning systems, and sustainability assessments increasingly depend on AI - but if models and training data remain proprietary, developing countries become consumers rather than co-creators. Open-source AI and open data are prerequisites for the inclusive, locally adapted solutions that the SDGs demand. Protection and promotion of human rights - AI systems deployed in environmental contexts - surveillance for conservation, predictive policing of resource extraction, automated permit decisions - carry significant human rights risks, particularly for Indigenous peoples, local communities, and environmental defenders. Ensuring that AI governance upholds due process, non-discrimination, and the right to a healthy environment is essential to preventing technology from deepening the inequalities it purports to address.

Environmental sustainability of AI systems and AI for planetary boundaries - The listed thematic areas do not explicitly address the environmental dimension of AI, which is both a cross-cutting enabler and an emerging risk. Critically, they also miss a foundational point: one of AI's core purposes must be helping humanity operate within planetary boundaries. If the most powerful technology of our era is not governed toward that goal, no amount of safe, trustworthy, or rights-respecting AI will matter on a destabilised planet. First, AI's material footprint is accelerating. Training and inference require vast energy, water, and critical mineral inputs. Data centre energy demand is projected to double by 2030, with associated carbon, water stress, and e-waste implications. Without governance frameworks that account for AI's lifecycle environmental costs, the technology risks undermining the very sustainability goals it is deployed to advance. Environmental impact assessment, disclosure standards, and circular design principles should be integrated into AI governance from the outset. Second, AI is already transforming environmental decision-making - from climate modelling and biodiversity monitoring to early warning systems and resource management. Yet there is no dedicated multilateral space connecting AI governance with the environmental treaty architecture (UNFCCC, CBD, UNEA, Basel/Stockholm/Minamata). The risk is parallel governance tracks that never converge: AI rules that ignore planetary boundaries, and environmental agreements that fail to harness or regulate AI. Third, the intersection raises distinct equity concerns. Countries bearing the heaviest environmental costs of AI infrastructure - mineral extraction, energy consumption, e-waste processing - often have the least capacity to benefit from AI or shape its governance. The Global Dialogue should treat environmental sustainability as a standing cross-cutting lens - similar to gender and human rights - and explicitly frame AI governance within the imperative of keeping human activity within safe ecological limits. The multilateral system's science-policy interfaces for environmental challenges offer directly transferable governance models.

Within the environmental sector, AI governance gaps are creating both acute risks and missed opportunities at a pace that existing institutions are not equipped to manage. The capacity asymmetry is widening. AI-powered environmental monitoring, including satellite imagery analysis, real-time emissions tracking, and biodiversity assessment, is advancing rapidly but almost entirely within a handful of technology companies and wealthy nations. Developing countries, which host the majority of the world's biodiversity and bear the greatest climate vulnerability, increasingly depend on proprietary AI systems they cannot audit, adapt, or afford. The absence of interoperable governance frameworks means there are no common standards for how AI-derived environmental data is validated, shared, or used in multilateral reporting, undermining trust in the very evidence base that treaty compliance depends on. Environmental decision-making is being automated without environmental governance. AI is already shaping resource allocation, land-use planning, conservation prioritisation, and climate risk assessment. Yet these deployments fall outside both AI governance discussions, which focus on general-purpose risks, and environmental treaty frameworks, which were not designed for algorithmic decision-making. This governance vacuum means consequential environmental choices are being made by systems with no transparency requirements, no environmental impact assessment, and no accountability to affected communities. The opportunity cost of inaction is compounding. Open AI models and open environmental data could dramatically accelerate progress on early warning systems, pollution monitoring, circular economy transitions, and nature-based solutions, particularly in regions where institutional capacity is thin. But without deliberate governance choices favouring openness, interoperability, and rights-based deployment, the default trajectory concentrates AI's environmental benefits among those who least need them. The most significant challenge is temporal: environmental tipping points and AI capability growth are both accelerating on exponential curves, while governance responses remain linear, fragmented, and siloed between the technology and environment communities.

The AI Dialogue has a unique opportunity to become the connective tissue between governance communities that are currently operating in parallel. Its greatest value lies not in creating new norms from scratch, but in bridging existing governance architectures with the emerging AI governance landscape. First, the Dialogue can serve as a translation layer between the technology and multilateral policy communities. Environmental treaties, human rights mechanisms, and trade frameworks have decades of governance experience with transboundary, exponential challenges, yet these lessons rarely reach AI policy discussions. Conversely, AI developers and governance practitioners often lack fluency in how multilateral processes actually function. The Dialogue can systematically surface transferable governance models, from science-policy interfaces like IPCC and IPBES to compliance mechanisms under the Montreal Protocol, and connect them to AI governance design. Second, the Dialogue can advance interoperability by mapping where AI governance intersects with existing international obligations. Countries are not starting from zero. They already have commitments under climate, biodiversity, chemicals, and human rights frameworks that AI deployment must respect. The Dialogue can help identify where AI-specific governance is genuinely needed versus where existing frameworks simply need updating to account for algorithmic decision-making. Third, the Dialogue can anchor international cooperation around shared goals rather than shared fears. Framing AI governance solely around risk mitigation misses the larger question: governance toward what? Keeping humanity within planetary boundaries, closing the digital divide, and accelerating the SDGs should be affirmative objectives that shape AI governance architecture, not afterthoughts. Finally, the Dialogue must model the inclusive, multistakeholder participation it advocates. This means meaningful engagement of developing countries, civil society, Indigenous peoples, and the environmental community, not as consulted observers but as co-designers of governance frameworks. The legitimacy of AI governance depends on it being shaped by those most affected, not only those most advanced.

The AI Dialogue should not duplicate what exists but instead become the platform that connects fragmented initiatives into a coherent governance ecosystem. Several existing mechanisms offer strong foundations. On the science-policy side, the AI Dialogue should build on the model established by IPCC, IPBES, and the International Resource Panel, which have decades of experience translating complex scientific evidence into policy-relevant assessments. The newly established International Scientific Panel on AI can learn directly from how these bodies handle uncertainty, achieve consensus across geopolitical divides, and maintain credibility with both scientists and policymakers. On governance architecture, the OECD AI Principles and the G7 Hiroshima Process have advanced normative frameworks among developed nations, but lack universal membership and developing country ownership. The Global Partnership on AI (GPAI) offers multistakeholder expertise but limited intergovernmental authority. The AI Dialogue can bridge this legitimacy gap by grounding these contributions within the universal UN framework while preserving their technical depth. On the environmental dimension, the Dialogue should connect with UNEP's digital transformation work, the Coalition for Digital Environmental Sustainability (CODES), and the science-policy processes under UNFCCC, CBD, and the chemicals and waste conventions. AI is already reshaping environmental monitoring, reporting, and compliance, yet none of these treaty processes have formal mechanisms to govern AI deployment within their domains. Regional initiatives also offer building blocks. The EU AI Act, the African Union's AI strategy, and ASEAN's governance frameworks represent diverse regulatory approaches that the Dialogue can help make interoperable rather than contradictory. The added value of the Dialogue is convening power and coherence. No other platform can simultaneously bring together AI developers, environmental treaty bodies, human rights mechanisms, and developing country governments. Its role should be to connect these communities, identify governance gaps between them, and ensure AI governance serves planetary sustainability, not just technological risk management.

The AI Dialogue's credibility will depend on whether it genuinely integrates diverse stakeholders into governance design or merely consults them as an afterthought. Format and structure should reflect this ambition. Governments should bring their existing governance experience, not just their AI strategies. Environment ministers, not only technology and foreign affairs ministries, need seats at the table. Countries with advanced environmental treaty implementation can offer transferable lessons on compliance mechanisms, science-policy interfaces, and common but differentiated responsibilities that directly apply to AI governance. The private sector, particularly AI developers and infrastructure providers, should contribute through mandatory transparency rather than voluntary commitments alone. This means disclosing environmental footprints of AI systems, sharing interoperability standards, and participating in governance design with accountability, not just advocacy. The Dialogue should create structured spaces where companies engage on governance terms set by the multilateral community, not the reverse. Civil society, Indigenous peoples, and affected communities bring the ground truth that technical and governmental stakeholders often lack. Their participation should be resourced and substantive, with speaking roles, drafting input, and the ability to challenge proposals, not simply observe proceedings. The scientific community, through the International Scientific Panel on AI and existing bodies like IPCC and IPBES, should provide regular evidence briefs that anchor discussions in empirical reality rather than speculation. On format, the Dialogue should avoid the trap of annual set-piece events that produce communiqués but no follow-through. Instead, it should adopt intersessional working groups on specific governance gaps, modelled on the contact groups and friends of the chair mechanisms used effectively in environmental negotiations. Thematic sessions should be co-designed by stakeholder categories rather than curated exclusively by governments. Outputs should be actionable: model governance frameworks, interoperability standards, and gap analyses that connect to existing treaty processes, not standalone declarations that sit alongside but never integrate with them.

Global AI governance discussions are dominated by technology ministries, AI companies, and a narrow band of civil society organisations concentrated in North America and Europe. Several critical communities remain structurally underrepresented. Environmental practitioners and treaty communities are largely absent from AI governance conversations, despite AI already reshaping environmental monitoring, compliance, and decision-making. The thousands of professionals working within UNFCCC, CBD, UNEA, and chemicals conventions have deep governance expertise but are rarely invited into AI policy spaces. Conversely, AI governance forums rarely engage with environmental evidence or planetary boundaries as a framing constraint. Developing country technologists and researchers, as distinct from developing country governments, bring practical knowledge of how AI is actually being deployed, adapted, and experienced in contexts of limited infrastructure, data scarcity, and institutional capacity. Their absence means governance frameworks are designed for conditions that do not reflect the majority of the world. Indigenous peoples and local communities are among the most affected by AI-driven environmental decisions, from conservation surveillance to resource extraction planning to land-use classification, yet have almost no presence in AI governance forums. Their knowledge systems and rights frameworks offer essential perspectives on what trustworthy and ethical AI means in practice. Workers in AI supply chains, from mineral extraction to data labelling to e-waste processing, experience the material costs of AI systems most directly but are invisible in governance discussions that focus on end-user impacts and developer responsibilities. Inclusion requires more than invitations. It demands resourced participation funds so that attendance is not limited to those with institutional travel budgets. It requires intersessional engagement mechanisms that allow input between formal meetings. It means multilingual processes and documentation as a default, not an accommodation. And it requires governance structures where underrepresented voices have drafting and decision-making roles, not just speaking slots in side events that no one with authority attends.

The AI Dialogue should break from the standard UN format of prepared statements delivered to near-empty rooms. If it is to govern a technology that moves at exponential speed, its engagement formats must match that ambition. Live governance simulations would be the single most valuable innovation. Rather than debating AI governance in the abstract, participants should work through real scenarios: an AI system misclassifies deforestation data used for carbon credit verification, an open-source climate model produces conflicting results across jurisdictions, an automated early warning system fails to alert a vulnerable community. These simulations force stakeholders to confront interoperability gaps, accountability questions, and equity dimensions in concrete terms rather than rhetorical ones. Structured red-teaming sessions should complement traditional panels. Invite AI developers to stress-test proposed governance frameworks in real time. Invite environmental treaty negotiators to identify where AI governance proposals conflict with existing multilateral obligations. This adversarial format produces sharper, more resilient outcomes than consensus-seeking plenaries. Asynchronous digital collaboration between formal meetings is essential. The environmental governance community has learned that governance happens between COPs, not only during them. The Dialogue should maintain open working documents, structured consultation platforms, and draft text repositories that stakeholders can contribute to continuously, not just during two days in New York. Reverse capacity-building sessions, where developing country practitioners present their AI deployment realities and governance needs to technology companies and developed country delegations, would invert the typical knowledge flow and ground discussions in lived experience rather than theoretical frameworks. Finally, the Dialogue should use AI itself as a governance tool within its own proceedings: real-time multilingual translation, AI-assisted synthesis of stakeholder submissions, and open dashboards tracking governance gaps and commitments. Demonstrating responsible AI use within the process builds credibility and practical understanding that no panel discussion can achieve.

Several existing governance models offer concrete, transferable solutions for AI governance rather than requiring the international community to design from scratch. The Montreal Protocol's Multilateral Fund demonstrates how to pair governance obligations with capacity support. It established that countries with fewer resources receive financial and technical assistance to comply with shared rules. An equivalent mechanism for AI governance could fund developing countries to build regulatory capacity, deploy open AI tools for public benefit, and participate meaningfully in governance processes rather than simply adopting frameworks designed elsewhere. The IPCC's assessment model shows how to build a credible science-policy interface for a complex, fast-moving domain. Its structured review process, confidence language, and government approval of summaries create legitimacy across geopolitical divides. The International Scientific Panel on AI should adopt these methods directly rather than reinventing them. The EU AI Act provides the most comprehensive attempt at risk-based AI regulation, classifying systems by their potential for harm and applying proportionate requirements. The African Union's AI Continental Strategy takes a different but equally instructive path, centering development priorities and local context over risk mitigation alone. A significant gap exists in tracking whether national AI and digital governance policies account for environmental sustainability. A new Digital Sustainability Policy Observatory, potentially hosted within the UN system, could systematically monitor all national AI, digital, and data governance frameworks for their environmental content, both positive provisions like green AI requirements and energy efficiency mandates, and negative gaps such as the absence of environmental impact assessment or lifecycle reporting obligations. This would create an evidence base for the AI Dialogue, enable peer learning, and make visible which countries are integrating planetary boundaries into their digital governance and which are not. Open-source AI initiatives like NASA's Earth science models demonstrate that high-capability AI for environmental applications can be built outside proprietary frameworks when governance choices deliberately favour openness.