Votee AI Limited

The first Global Dialogue on AI Governance represents a critical juncture for humanity. As the UN Secretary-General aptly noted, the question is no longer whether AI will transform our world, but whether we will govern this transformation together. In my view, the success of this inaugural Dialogue hinges on three key outcomes, establishing human-centric accountability, protecting cultural and emotional authenticity, and ensuring linguistic equity. First, the Dialogue must move beyond abstract ethical principles to operationalize the "human-in-the-loop" mandate across all sectors. As a scientist, I believe AI should augment rather than replace human agency. Success means forging a global consensus that humans must bear ultimate responsibility for AI-assisted decisions, ensuring rigorous verification and maintaining the integrity of our pursuit of truth. Second, the Dialogue must act as an interdisciplinary translator, elevating voices beyond the technologists and policymakers of the Global North. My experience in the performing arts underscores the need to safeguard spaces for unmediated human creativity. A successful outcome would involve frameworks that protect the emotional resonance of human expression and prevent the homogenization of culture by algorithmically optimized outputs. Finally, a truly successful Dialogue must establish a linguistic equity floor. Currently, AI systems perform poorly for billions of people whose native languages are neglected by global models. Governance must demand that AI systems serving public-interest functions demonstrate measurable competence in the languages of the jurisdictions they serve. By integrating the rigorous verification demands of science with the expressive imperatives of the arts, the Dialogue can create a holistic governance model. Success will not be measured by formal statements alone, but by dynamic, continuous collaboration that ensures AI remains a supportive tool that respects the unique ideas and inherent dignity of every human being.

• Transparency, accountability, and human oversight
• Social, economic, ethical, cultural, linguistic and technical implications of AI
• Safe, secure and trustworthy AI
• Protection and promotion of human rights

As a scientist, my perspective is grounded in the principle that AI should augment, not replace, human agency and creativity. I prioritize transparency, accountability, and human oversight because, while absolute certainty is unattainable in science, rigorous verification is indispensable. Understanding how AI systems reach their conclusions is essential to preserving the integrity of truth-seeking, and ultimate responsibility must remain with humans. My background in the performing arts further shapes this view. Art reflects human experience, our social, cultural, linguistic, and ethical realities. AI therefore carries implications far beyond the technical domain. It must not homogenize expression or dilute the emotional depth that arises from authentic human experience. Preserving diversity in voice and meaning is as critical as ensuring technical performance. For these reasons, I emphasize the importance of safe, secure, and trustworthy AI. Robust governance frameworks are necessary to ensure that AI systems support human endeavours without introducing unacceptable risks, whether in scientific research or public discourse. Safety is not merely a technical requirement but a societal obligation. Equally important is the protection and promotion of human rights. As AI becomes embedded in daily life, we must safeguard individual rights to self-expression, intellectual property, and autonomy. Meaningful collaboration, whether in a laboratory or on a stage-depends on respecting the originality, dignity, and responsibility of each individual.

An emerging issue is the preservation of epistemic humility and authentic human connection in an increasingly automated world. While the listed themes address structural and ethical governance, they overlook the psychological and philosophical impact of AI on human identity and our collective pursuit of meaning. In science, the danger is not just that AI might be wrong, but that humans might become overly reliant on it, losing the critical thinking skills and the collaborative friction that sparks true innovation. The ideas generated during spontaneous human conversation, the "chatting" where real breakthroughs often occur. It cannot be fully replicated by a machine. In arts, the issue is the potential devaluation of human emotion and lived experience. Art is a conduit for conveying profound messages to an audience, creating resonance based on shared human struggles and triumphs. If AI-generated content floods the cultural space, we may feel a deep sense of alienation, where the "who I am" is overshadowed by algorithmically optimized outputs. Therefore, a cross-cutting theme must be the safeguarding of human-centric processes, ensuring that AI governance actively protects spaces for unmediated human creativity, emotional expression, and the irreplaceable value of human-to-human collaboration.

Across Hong Kong and the broader Asia-Pacific region, the most pressing AI governance gap is twofold. First, leading global models remain functionally inadequate in many regional languages. Research indicates that AI accuracy can drop to around 50% for non-English contexts, with even poorer performance in code-switched, low-resource, and culturally nuanced settings such as Cantonese. Second, heavily regulated sectors, including banking, healthcare, and public administration, face structural barriers in adopting these systems, as black-box models cannot meet stringent audit, traceability, and compliance requirements. This gap contributes to a deeper challenge, namely, the erosion of trust and accountability. In scientific domains, the rapid deployment of opaque AI systems undermines the integrity of scientific methods. When researchers cannot verify how models generate hypotheses or interpret data, the risk of untraceable errors increases, potentially leading to flawed or irreproducible findings. Similarly, in the arts, the absence of clear ethical frameworks for generative AI raises concerns over intellectual property and creative ownership. It also diminishes the emotional authenticity of artistic expression when audiences struggle to distinguish between machine-generated outputs and genuine human experience. Yet these challenges also create an opportunity to redefine human value. Acknowledging that AI systems are inherently imperfect reinforces the importance of human oversight, rigorous peer review, and collaborative validation in science, elevating researchers from data processors to critical evaluators. In the arts, the rise of AI sharpens the focus on uniquely human qualities, emotion, lived experience, and personal responsibility, positioning AI as a supportive tool rather than a substitute for human creativity.

The AI Dialogue can connect disparate disciplines, cultures, and governance models, functioning much like the collaborative conversations that drive both scientific discovery and artistic creation. First, it can establish a global baseline for human oversight and accountability. By bringing nations together, the Dialogue can forge consensus that AI must remain a subordinate tool to human agency, ensuring that individuals and institutions ultimately bear responsibility for AI-assisted outcomes. Second, it can act as an interdisciplinary translator. The implications of AI are not merely technical; they are deeply social and cultural. The Dialogue can elevate the voices of artists, philosophers, and social scientists alongside technologists and policymakers, ensuring that governance frameworks protect the emotional and cultural resonance of human societies, not just economic interests. Third, it can promote shared standards for transparency and verification, which are essential for global scientific collaboration. By fostering open discussions on how to audit and interpret AI systems, the Dialogue can help the international scientific community maintain rigorous standards of truth-seeking, acknowledging AI's fallibility while safely harnessing its analytical power. Fourth, the Dialogue can establish a linguistic equity floor for AI governance, a shared expectation that AI systems serving public-interest functions in any jurisdiction must demonstrate measurable competence in that jurisdiction's languages, including those currently neglected by global models. Without such a floor, the asymmetry between language communities will harden into a structural feature of the AI era.

The AI Dialogue should build upon existing frameworks that prioritize human rights, scientific integrity, and cultural preservation. Key initiatives include the UNESCO Recommendation on the Ethics of Artificial Intelligence, which provides a strong foundation for the cultural and ethical implications of AI, and the OECD AI Principles, which emphasize transparency, robustness, and accountability. Furthermore, connecting with international scientific bodies like the International Science Council (ISC) and Global Artists Coalition is essential. While existing mechanisms often operate in silos focusing either on economic development, technical standards, or abstract ethics, the AI Dialogue can bring synthesized, actionable governance. Its primary added value would be operationalizing the concept of "human-in-the-loop" across diverse sectors. For scientists, it can translate abstract transparency principles into concrete guidelines for peer-reviewing AI-assisted research. For artists, it can bridge the gap between intellectual property law and ethical AI training, ensuring that the human emotional experience remains the core of cultural production. By integrating the rigorous verification demands of science with the expressive imperatives of the arts, the AI Dialogue can create a holistic governance model that truly reflects the complexity of human identity and responsibility.

To harness collective human wisdom, the AI Dialogue must move beyond rigid diplomacy and adopt dynamic, collaborative formats, making it closer to scientific symposia or artistic workshops. It should emphasize genuine exchange, co-creation, and iterative problem-solving rather than formal statements alone. Scientists and researchers must ground the Dialogue in evidence, offering clear assessments of AI's capabilities and limits while sharing methods to verify outputs and uphold the Scientific Method. Artists and cultural practitioners add a critical lens, expressing AI's emotional and ethical impacts beyond technical language and defending authentic human expression. Technologists, in turn, must commit to transparency, explaining how systems work, where they fail, and how to improve them, while helping design practical accountability mechanisms. No single group has all the answers. Value emerges from their convergence. The Dialogue should be structured around interdisciplinary working groups defined by shared challenges, such as "AI and the Nature of Truth," rather than professional silos. Informal, off-the-record spaces are also essential, as candid conversations often produce the most original ideas. An often-missing voice is deployment-grade infrastructure providers, those operating AI in regulated sectors like banking, healthcare, and public administration. Their experience with audit, procurement, and data governance constraints provides practical insights distinct from research or policy perspectives. Finally, the Dialogue must be continuous, not one-off—testing and refining governance frameworks through real-world feedback in an iterative, transparent process open to revision.

Global discussions on AI governance are dominated by technologists, corporate leaders, and policymakers from the Global North. This concentration of influence creates a structural imbalance: perspectives that focus on efficiency and scale are prioritized, while those addressing meaning, culture, and human experience are often excluded. As a result, voices from the arts, humanities, Indigenous communities, and independent scientists remain underrepresented, limiting a holistic understanding of AI's societal impact. This imbalance becomes clearer when examining how different forms of knowledge are treated. Artists and humanities scholars explore emotion, belief, and cultural context, dimensions increasingly shaped by AI but rarely reflected in technical or regulatory frameworks. At the same time, independent and academic scientists who emphasize careful validation and methodological rigor often receive less attention than well-funded industry labs focused on speed and deployment. Non-Western philosophical traditions, particularly those grounded in oral heritage and ecological relationships, are also marginalized by data-driven systems that privilege measurable inputs over contextual knowledge. The most fundamental gap, however, lies in language. Nearly three billion people use languages for which AI systems perform poorly, sometimes with accuracy as low as 50%. Their needs are largely interpreted by Global North intermediaries rather than addressed through systems built in their native linguistic and cultural contexts. This is not just a representation issue, but an infrastructure gap. Closing these gaps requires structural change. Decision-making bodies must directly include cultural experts, ethicists, and independent researchers. Engagement methods should expand beyond formal policy to include storytelling and localized knowledge. Most importantly, resource equity—funding, infrastructure, and translation—is essential to enable sustained participation and the development of truly inclusive AI systems.

AI governance dialogues should move beyond rigid, formal structures and adopt more human-centred formats that engage both analytical and emotional intelligence. Traditional discussions often remain abstract; more immersive approaches can help participants better understand real-world consequences and foster empathy. One approach is performance-based engagement, such as theatrical or immersive storytelling that simulates the societal impact of AI policies. By making outcomes tangible, participants can grasp implications that are difficult to capture through technical debate alone. Complementing this, interdisciplinary policy hackathons can bring together scientists, artists, lawyers, and engineers to co-create policy prototypes within short timeframes. This format encourages rapid idea generation and productive tension across perspectives, leading to more creative and grounded solutions. Informal dialogue formats are equally important. Fireside-style conversations, without formal presentations, create space for active listening, reflection, and honest discussion. Such settings make it easier for participants to express uncertainty and acknowledge the limits of their knowledge, which is essential for responsible governance. Finally, governance frameworks should be validated through multilingual policy testbeds. Standards such as transparency, watermarking, and auditability must be stress-tested in real-world deployments, especially in low-resource and neglected-language environments. Frameworks that perform well in English often fail elsewhere, such as in low resource languages, revealing hidden assumptions only at deployment. Embedding this testing phase ensures that AI governance is not only theoretically sound but also globally robust and inclusive.

Effective AI governance must balance innovation with strong human oversight. A core mechanism is the "human-in-the-loop" mandate, requiring human review for high-stakes decisions such as medical diagnoses, criminal justice outcomes, and scientific publication. This ensures accountability remains with humans and recognizes that AI systems are inherently fallible. Technical safeguards should complement this oversight. Watermarking and provenance standards, such as those developed by the Coalition for Content Provenance and Authenticity, embed metadata into digital content, enabling clear distinctions between human-created, authentic, and AI-generated outputs. These measures help preserve trust and protect the integrity of both scientific and artistic work. Another key tool is the use of Algorithmic Impact Assessments (AIAs). Like environmental assessments, AIAs should be mandatory before deploying systems in sensitive domains, requiring developers to evaluate social, ethical, and cultural risks. This ensures innovation aligns with human values rather than undermining them. For regulated sectors such as finance, healthcare, and public administration, governance must go further. High-stakes systems should require on-premise, auditable deployment with zero data leakage and full provenance tracking. In such contexts, compliance depends not on promises but on verifiable, jurisdiction-specific evidence, effectively turning data sovereignty into an audit requirement. Finally, open science and reproducibility are essential, especially in research. Requiring access to underlying models and data enables independent verification and strengthens transparency. This reinforces a core scientific principle: trustworthy knowledge must be testable, even in the presence of uncertainty.