AI Ethics Lab, Rutgers University

A successful first Global Dialogue on AI Governance would help translate shared commitments into actionable frameworks for governing AI, not as a tool but as an environment. Humans inhabit and shape this environment, and it shapes us in return. This raises profound questions about what it means to exercise human agency and expression in an age of co-intelligence. A successful Dialogue would reflect this shift, recognizing that governing AI requires stewarding an ecosystem, not simply regulating a tool. Second, success would include operationalizing the global commitment to human dignity. Our research at the AI Ethics Lab at Rutgers University shows that 82% of national constitutions and 52 multinational instruments explicitly reference dignity. Our Dignity by Design model shows how to embed that commitment across the AI lifecycle, from data collection to training, deployment, and monitoring. Third, the Dialogue should emphasize the importance of reducing moral distance. Many AI-related harms arise from cultural, physical, psychological, and procedural distance between those who build and regulate AI systems and those whose lives they affect. Cultivating moral imagination, the ability to understand multiple perspectives and see the human impact of decisions, is essential to building trustworthy AI environments. Finally, success would include applying a vulnerability lens that brings attention to those most susceptible to harm, including Indigenous Peoples, identity groups, at-risk populations, and displaced people.

• Protection and promotion of human rights
• Transparency, accountability, and human oversight
• Interoperability of governance approaches
• Social, economic, ethical, cultural, linguistic and technical implications of AI

Protection and promotion of human rights is central because the global commitment to preserve the inherent worth and dignity of every person must guide how AI environments are built and regulated. Transparency, accountability, and human oversight are essential because recommendations can become prescriptions. Over time, people choose from what AI generates, which raises profound questions about what it means to be human in an age of co-intelligence. This requires clear responsibility across the AI lifecycle, from data collection to training, deployment, and monitoring. Interoperability of governance approaches matters because AI operates across cultures, regions, and legal systems. This is why we ask not only which rights are at stake, but what cultural contexts and regional approaches to human rights are used to interpret these risks. Finally, the social, economic, ethical, cultural, linguistic, and technical implications of AI are central because AI is not merely a technical tool. It is an environment that humans inhabit and shape, and that shapes us in return. In this environment, we are not passive recipients of technological change. We are co-architects, with a duty to shape AI environments that reduce harm and benefit all.

One cross-cutting issue is the problem of moral distance. Many of the AI-related harms we are seeing are not due to a lack of intelligence. They are failures of imagination. The distance can be cultural, physical, psychological, or procedural. Each stage of the AI lifecycle can widen the moral gap between those who build and regulate AI systems and the people whose lives they affect. To address this, we need to cultivate moral imagination, the ability to understand multiple perspectives, and to see the human impact of technological decisions. This is essential to building trustworthy AI environments. A second issue is how we understand AI itself. If AI is treated as a tool, people can be reduced to data providers. If AI is understood as an environment, then people are participants who shape it and are shaped by it in return. That shift changes how we think about responsibility, participation, and benefit. Third, we must apply a vulnerability lens. Some populations are more susceptible to harm and require greater attention and protection. In our work at the AI Ethics Lab, this includes Indigenous Peoples, identity groups, at-risk populations, and displaced people. Those most affected must have a meaningful say in the systems that shape their lives. Finally, we should ask whether "smart" tech is the right goal. The real question is whether we are developing the capacity to make wiser choices to ensure the century's technological advancements support human flourishing across cultures and generations.

AI governance gaps are affecting our sector in how AI recommendations guide human choices, even as the options presented are filtered through models trained on dominant cultures. This creates real challenges for cultural rights and the right to development, especially when people come to expect repeated guidance and begin choosing from what AI generates. One of the most significant challenges is that the options people see are shaped by dominant cultural datasets. This raises concerns about whose knowledge, creativity, and perspectives are being represented, and whose are being filtered out. Another challenge is ensuring equitable participation in the benefits of technological progress. The right to development is not only about preventing harm. It is also about whether the communities reflected in the data share in the benefits of the technologies built from it. At the same time, there are important opportunities. AI governance can be approached as stewarding an ecosystem, comparable to the cooperative governance of global financial markets. This opens the possibility for shared responsibility and coordinated action across sectors and regions. There is also an opportunity to strengthen participation. The most vulnerable must be invited to participate in, contribute to, and enjoy economic, social, cultural, and political development. It is that meaningful participation that makes AI systems trustworthy. Finally, governance can move beyond efficiency, automation, and optimization toward a broader vision that balances individual, community, and environmental interests over the short and long term, ensuring that technological advances reduce harm and actively maximize good for all.

The AI Dialogue can play a central role by helping translate a shared legal foundation into a continuous, cooperative governance system. Our work at the AI Ethics lab rests on the premise that dignity is the status on which all rights are grounded. Legal systems express dignity in different ways, but the larger pattern is clear: dignity explains why the human being has standing and why rights must be respected, protected, and fulfilled. This provides a foundation for international cooperation. The AI Dialogue can advance this by supporting Dignity by Design as a foundational practice embedded at every stage of the AI lifecycle and as a central professional practice for monitoring and regulating AI systems. Building on this legal foundation, cooperation can move from a legal index to a technical protocol to a governance flywheel. The Protocol translates human rights into a structured technical system. CHARTER, Classifying Human Rights Advancements for Responsible Technology and Ethical Response, then serves as a continuous feedback system. Expert AI trainers from around the world apply structured rubrics. Real-world evidence is gathered from audits, NGO reports, academic research, and cross-cultural perspectives. Frameworks are revised, models are aligned, and updated versions are released. This process enables ongoing evaluation and shared governance across regions. It also reflects a Duty of Care by bringing attention to populations at heightened risk and ensuring that governance is informed by cross-cultural perspectives. In this way, the Dialogue can support cooperation that is not static, but iterative, participatory, and grounded in human dignity.

The AI Dialogue should build upon the existing body of international human rights law and the institutional frameworks that interpret and apply those rights across regions. The AI & Human Rights Index draws on eight decades of human rights law, including declarations, conventions, covenants, and multinational frameworks, to organize the relationships among rights, instruments, principles, sectors, and vulnerable populations. This includes work developed across the United Nations system and regional bodies. These instruments already provide a legal taxonomy for evaluating how AI systems may violate or advance human rights. The added value of the AI Dialogue is to connect these legal frameworks to technical and governance systems. The Protocol provides a semantic infrastructure that translates this legal framework into a structured system for evaluating AI. SKOS and OWL are used to organize concepts, define relationships, and support technical implementation, while recognizing that human rights are continually contested and reinterpreted. CHARTER then builds on this by creating a governance flywheel. Expert AI trainers apply structured rubrics, gather real-world evidence from audits, NGO reports, academic research, and cross-cultural perspectives, and continuously revise and update frameworks. The added value of the AI Dialogue is to bring these elements together globally, enabling shared evaluation, continuous refinement, and coordinated governance. It can also strengthen the Duty of Care by prioritizing populations at heightened risk and supporting governance systems informed by diverse cultural and regional perspectives.

Different stakeholders can contribute by participating in a structured process that moves from a legal framework to a technical system to a continuous governance cycle. The AI & Human Rights Index organizes the relationships among rights, instruments, principles, sectors, and vulnerable populations. This provides a shared framework that can be used across sectors to evaluate how AI systems may violate or advance human rights. Stakeholders can contribute by applying structured rubrics to evaluate AI systems, drawing on their expertise in law, technology, and sector-specific practices. This includes participation from governments, companies, academics, and civil society. Expert AI trainers from around the world can apply these structured evaluations through reinforcement learning from human feedback. Real-world evidence can then be gathered from audits, NGO reports, academic research, and cross-cultural perspectives. This evidence is used to revise frameworks, align models, and release updated versions. In this way, participation is not a one-time input, but part of a continuous feedback system. The structure of the Dialogue can reflect this process: shared frameworks, structured evaluation, cross-cultural input, and ongoing revision. This supports governance that is iterative, participatory, and grounded in human rights.

Some populations are more susceptible to harm and require greater attention and protection. This includes Indigenous Peoples, identity groups, at-risk populations, and displaced people. These communities are often underrepresented in global discussions, even though they are among those most affected by AI systems. Applying a vulnerability lens is essential. This lens brings attention to populations at heightened risk and ensures that their experiences are central to evaluating how AI systems may violate or advance human rights. Inclusion requires meaningful participation. The most vulnerable must be invited to participate in, contribute to, and enjoy economic, social, cultural, and political development. This can be supported by engaging stakeholders from different cultures, regions, and professional backgrounds. Cross-cultural perspectives should be incorporated into the evaluation process through structured input, real-world evidence, and ongoing feedback. By applying this approach, governance can move from marginalization to mattering, ensuring that those most affected have a meaningful say in the systems that shape their lives.

Innovative engagement can be grounded in a continuous feedback system that connects legal frameworks, technical systems, and real-world evaluation. The AI & Human Rights Governance Flywheel, CHARTER, provides one such model. Expert AI trainers from around the world apply structured rubrics through reinforcement learning from human feedback. This allows diverse stakeholders to participate directly in evaluating AI systems. Engagement can also include the collection of real-world evidence from audits, NGO reports, academic research, and cross-cultural perspectives. This creates a dynamic process where input is continuously integrated and used to revise frameworks. Interactive tools can support this process by helping participants apply structured frameworks to specific claims and case studies. This allows stakeholders to move from abstract discussions to concrete evaluation. These formats support engagement that is not static, but iterative. Participants contribute to ongoing evaluation, refinement, and governance of AI systems in relation to human rights. In this way, engagement becomes a process of shared learning, structured input, and continuous improvement across sectors and regions.

One effective approach is Dignity by Design. Our research at the AI Ethics Lab shows that dignity is the status on which all rights are grounded. Legal systems express dignity as a principle, as a right, or as both, but the larger pattern is that dignity explains why the human being has standing and why their rights must be respected, protected, and fulfilled. Dignity by Design identifies where this commitment must be embedded across the AI lifecycle, from data collection and system design to training, deployment, and monitoring. This includes asking not only whether data is lawfully collected, but whether communities share in the benefits of the technologies built from it. A second approach is to apply a vulnerability lens through what we call the Duty of Care protocol. Some populations are more susceptible to harm and require greater attention and protection, including Indigenous Peoples, identity groups, at-risk populations, and displaced people. This approach draws attention to the most vulnerable and ensures they are invited to participate in, contribute to, and enjoy development. A third practice is moving from moral distance to moral imagination. Cultural, physical, psychological, and procedural distance can separate those who build and regulate AI systems from those whose lives they affect. Cultivating moral imagination enables people to understand multiple perspectives and see the human impact of technological decisions, reducing that distance. Finally, the Governance Flywheel, CHARTER, provides a concrete mechanism for implementation. Expert AI trainers apply structured rubrics, gather real-world evidence from audits, NGO reports, academic research, and cross-cultural perspectives, and continuously revise and update frameworks. This creates an ongoing process of evaluating, refining, and governing AI systems in relation to human rights.