Xiamen University Malaysia

A successful first Global Dialogue on AI Governance should be judged by whether it produces useful cooperation, not only high-level statements. For me, the most important outcome would be a shared understanding among countries about the main risks and opportunities of AI, especially where national approaches may differ. It should also give developing countries a stronger voice. Many countries are expected to manage the effects of AI without the same technical, financial, or regulatory capacity as larger economies. A successful dialogue should therefore identify practical support for policy design, public-sector training, research capacity, and access to safe and beneficial AI tools. Another important outcome would be agreement on a few common priorities, such as protecting human rights, reducing discrimination, improving transparency, supporting innovation, and preventing misuse. These priorities should be simple enough to guide action, but flexible enough for different national contexts. The dialogue should also help reduce fragmentation. Countries will not adopt identical AI laws, but they can work toward compatible standards, shared terminology, common risk-assessment approaches, and mutual learning from existing governance models. The first dialogue should create a clear path for follow-up. It should not end as a one-time meeting. Success would mean clear next steps, regular reporting, continued participation from governments and non-government actors, and links with scientific and technical expertise. I would consider the first Global Dialogue successful if it builds trust, gives underrepresented countries a real role, and turns broad concern about AI into practical cooperation.

Safe, secure and trustworthy AI;Open-source software, open data and open AI models

I selected safe, secure and trustworthy AI because any global discussion on AI governance must begin with public trust. AI systems are increasingly used in education, health, employment, public services, and decision-making. If these systems are unsafe, biased, insecure, or difficult to understand, they can cause real harm and reduce confidence in innovation. Strong governance should therefore support transparency, accountability, risk assessment, security, and human oversight. I also selected open-source software, open data and open AI models because access is a major global issue. Many countries, researchers, small companies, and public institutions do not have the same resources as large technology firms. Open tools can help reduce this gap by supporting local innovation, independent research, education, and public-interest applications. At the same time, openness must be managed carefully. Open models and datasets can create benefits, but they can also be misused or reproduce harmful biases if they are released without proper safeguards. For this reason, I see these two areas as closely connected. The goal should not be openness without responsibility, or safety without access. A successful governance approach should find a balance: enabling broad participation in AI development while ensuring that systems are secure, reliable, fair, and aligned with human rights.

1:Power concentration 2:Environmental impact of AI 3:Labour and employment disruption 4:Misinformation and public trust 5:Digital dependency Power concentration is a major concern. Advanced systems need large amounts of data, computing power, money, and skilled workers. This can leave too much influence with a few large companies and powerful countries. Environmental cost should also be considered. Running large systems can use high levels of energy, water, and hardware. Governance discussions should consider whether these costs are fair and sustainable. Public benefit needs stronger attention. New technologies should support health, education, agriculture, climate response, disaster management, and public services, especially in communities with fewer resources. Digital dependency is another risk. Many countries may rely on foreign platforms, cloud services, and private tools. This can weaken local control, data sovereignty, and long-term national capacity. Work and employment also need attention. Automation may change job roles, increase workplace monitoring, and create pressure for new skills. Workers, students, and education systems need support to adapt. Misinformation and trust are serious concerns. Synthetic text, images, audio, and video can affect elections, public debate, science communication, and social stability. Practical measures are needed to identify harmful content and protect public trust.

In Malaysia and the wider Southeast Asian region, governance gaps around safe and trustworthy systems are creating both concern and opportunity. Many public and private organisations are adopting new tools faster than rules, training, and oversight can develop. This creates risks in areas such as education, health, finance, hiring, public services, and research, where errors, bias, privacy breaches, or unclear accountability can affect real people. A major challenge is uneven capacity. Larger institutions and companies can access better infrastructure, technical expertise, and legal support, while smaller universities, public agencies, start-ups, and community organisations may struggle to evaluate systems properly. This can widen existing gaps between urban and rural areas, large and small institutions, and high- and low-resource sectors. Open-source software, open data, and open models create an important opportunity for countries like Malaysia. They can support local research, local-language tools, education, innovation, and public-interest applications without complete dependence on expensive foreign platforms. This is especially important for Bahasa Melayu, indigenous languages, regional health needs, agriculture, disaster response, and small-business development. However, openness also brings risks when datasets are poor quality, culturally narrow, biased, or released without safeguards. Open models may also be misused for fraud, misinformation, cyber abuse, or academic misconduct. Therefore, the main need is not to block openness, but to combine it with clear safety practices, documentation, testing, responsible release, and human oversight. The biggest opportunity is to build a balanced governance approach: one that protects people while still allowing local innovation. For the education and research sector, this means clear institutional policies, training for staff and students, support for responsible open science, and stronger regional cooperation. Governance done well can help Malaysia benefit from these technologies rather than remain only a consumer of tools built elsewhere.

The AI Dialogue can help countries work together on issues that no country can manage alone. AI systems are developed, sold, and used across borders, so weak governance in one place can create risks elsewhere. A regular global forum can help countries share experience, compare policy approaches, and learn from both successes and mistakes. Its role should be to build common ground without forcing every country to follow the same model. Countries differ in legal systems, resources, languages, cultures, and development needs. The Dialogue can support shared understanding on safety, accountability, transparency, human oversight, and responsible use, while still respecting national context. It can also reduce confusion between different governance approaches. As more countries create their own rules, there is a risk of fragmented standards. The Dialogue can help identify compatible practices, common terms, and basic expectations for high-risk uses. A strong contribution would be support for countries with limited technical and regulatory capacity. Many developing countries need help with policy design, regulator training, public-sector readiness, research infrastructure, and access to reliable tools. The Dialogue can connect them with expertise and partnerships. The Dialogue should also include voices beyond governments, including universities, civil society, industry, technical experts, and affected communities. This would make cooperation more realistic and more trusted. Its success will depend on whether discussion leads to practical steps, such as shared guidance, capacity-building, evidence exchange, and regular follow-up.

The AI Dialogue should connect with existing work rather than start from zero. Useful foundations include the UNESCO ethics recommendation, the OECD AI Principles, the Global Partnership on AI, the G7 Hiroshima process, the AI Safety Summit process, ITU's AI for Good work, and regional efforts such as ASEAN's guidance on AI governance. These initiatives already offer principles, policy experience, technical discussions, and examples of cooperation. They cover areas such as human rights, safety, innovation, standards, capacity-building, and responsible use. The Dialogue can learn from them and avoid repeating the same discussions. Its added value would be wider participation. Some existing initiatives are led mainly by advanced economies, technical groups, or smaller coalitions. A UN-based Dialogue can give all countries a more equal place at the table, including developing countries that are often affected by AI but have less influence over its rules. The Dialogue can also help connect separate efforts. At present, many initiatives work in parallel. This can create overlap, confusion, and gaps. The Dialogue can compare approaches, identify common ground, and support more compatible governance without forcing one global rulebook. It can also link policy discussion with scientific and technical evidence, especially through the International Independent Scientific Panel on AI. This would help countries make decisions based on evidence rather than pressure from powerful actors. The main added value should be practical: shared guidance, capacity-building support, knowledge exchange, regional cooperation, and regular follow-up. In this way, the Dialogue can act as a bridge between existing initiatives and a more inclusive global governance process.

Governments can share national experiences, legal gaps, public-sector needs, and priorities for cooperation. They should also explain what support they need to govern AI safely. Industry can provide technical knowledge, risk information, and practical experience from developing and deploying systems. They should also be expected to explain safety practices, limitations, and accountability measures. Academia and researchers can contribute independent evidence, evaluation methods, and critical analysis. Their role is important for separating real risks and benefits from hype. Civil society can represent public concerns, human rights, affected communities, workers, children, and marginalized groups. This will help keep the Dialogue connected to real social impact. Technical standards bodies can help translate broad principles into usable practices, such as testing, documentation, audit methods, and risk assessment. The format should be simple and action-oriented. The Dialogue could include a high-level plenary for political direction, smaller working groups for specific issues, and regional sessions so that local needs are not lost. Each session should have a clear question, a short background note, and a written outcome. The structure should avoid long speeches only. It should include case studies, problem-solving sessions, and space for countries to share concrete needs. Developing countries should have funded participation where needed, so inclusion is practical, not only symbolic. There should also be a public summary after each Dialogue, with agreed priorities, unresolved issues, and next steps. Between meetings, working groups could continue online and report progress. This would make the Dialogue more useful, transparent, and continuous.

Developing countries need stronger representation, especially small states and low-resource countries that may use imported tools without much influence over how they are designed or regulated. Local-language communities are often missing. Many systems work better in dominant global languages, while smaller languages, indigenous languages, and regional dialects receive less attention. Affected communities should also be heard directly. This includes workers, students, patients, migrants, women, children, persons with disabilities, and communities exposed to surveillance, automated decision-making, or online harm. Small businesses and public institutions are also important. They often adopt new tools without the legal, financial, or technical support available to large companies. Independent researchers and civil society groups need better access to discussions, data, and evidence. They can identify risks and social impacts that may be overlooked by governments or industry. These groups could be included through regional consultations, funded participation, multilingual engagement, open calls for written input, and community-level listening sessions. The Dialogue should not depend only on high-level panels. It should create spaces where people can explain real problems from their own setting. There should also be transparent reporting showing whose input was received and how it shaped the final outcomes. This would make participation more meaningful and build trust in the process.

Problem-solving roundtables would allow participants to work on real governance challenges, such as bias in public services, misuse of open models, or lack of regulator capacity. This would be more practical than general speeches. Regional listening sessions could bring in views from Africa, Asia-Pacific, Latin America, small island states, and other underrepresented regions before the main meeting. Their findings should feed directly into the agenda. Case-study clinics could examine real examples from health, education, elections, labour, and public services. Participants could discuss what worked, what failed, and what lessons can be shared. Youth and community panels would help include people who are often affected by technology but rarely shape policy. These sessions should be structured around lived experience, not technical language only. Multi-stakeholder working groups could continue between meetings and produce short outputs, such as guidance notes, checklists, or policy options. Scenario exercises would also be useful. Participants could respond to realistic situations, such as a harmful automated decision, a deepfake crisis, or a cross-border data problem. This can reveal gaps in current governance. Open written submissions and multilingual online consultations would allow wider participation from people who cannot attend in person.

UNESCO Recommendation on AI Ethics This gives a global ethical framework and includes a Readiness Assessment Methodology to help countries identify legal, institutional, and capacity gaps. It is useful because it connects principles with country-level diagnosis and support https://www.unesco.org/en/artificial-intelligence/recommendation-ethics OECD AI Principles and OECD.AI Policy Observatory The OECD Principles give policy guidance on trustworthy AI, while the OECD.AI platform collects national AI policies and initiatives. This can help countries compare approaches and avoid starting from zero. https://www.oecd.org/en/topics/sub-issues/ai-principles.html NIST AI Risk Management Framework The NIST framework offers a practical way for organisations to identify, measure, manage, and govern AI risks. It is useful because it translates broad values into operational risk-management steps. https://www.nist.gov/itl/ai-risk-management-framework ASEAN Guide on AI Governance and Ethics For Southeast Asia, this is especially relevant. It offers practical guidance for organisations and supports alignment across ASEAN while respecting regional needs. https://asean.org/wp-content/uploads/2024/02/ASEAN-Guide-on-AI-Governance-and-Ethics_beautified_201223_v2.pdf National AI offices and strategies Malaysia's National AI Office is one example of a national mechanism intended to coordinate policy, regulation, research, and strategic planning. Such bodies can connect global dialogue with domestic implementation. https://www.reuters.com/technology/artificial-intelligence/malaysia-launches-national-ai-office-policy-regulation-2024-12-12/