Governments around the world are increasingly aligning around a shared set of aspirations for artificial intelligence: it should be safe, transparent, accountable, trustworthy, and centered on people. The greater challenge is turning those aspirations into requirements that can be implemented, tested, and recognized across countries and industries.
During the session From principles to practice: The role of international standards in AI at the AI for Good Global Summit 2026, the leaders of the International Organization for Standardization (ISO), the International Electrotechnical Commission (IEC), and ITU’s Telecommunication Standardization Bureau examined how standards can bridge that implementation gap.
Moderated by journalist Charlotte Kan, the discussion explored how standards create a common language for AI governance, help translate regulation into operational processes, reduce fragmentation, and enable more countries and stakeholders to shape the systems that will affect them.
Creating a common language for AI governance
Sergio Mujica, Secretary-General of ISO, began by clarifying the relationship between policy, regulation, and standards.
“We do not create policies, we do not establish regulations. That’s the job of policymakers and governments. What we do is support the implementation of those ambitions,” Mujica said.
This supporting role involves translating broad ambitions into agreed definitions, processes, technical requirements, and methods of verification.
It becomes particularly important because AI governance is often developed nationally, while AI systems, services, and supply chains operate across borders. Different regulatory approaches can therefore create fragmentation unless countries have compatible ways of interpreting and applying them.
Standards can help connect these frameworks by establishing a shared vocabulary. Terms such as risk, impact assessment, accountability, and AI itself may appear across many policies, but they must be understood consistently if different organizations and jurisdictions are to work together. They can also provide processes that governments can integrate into different regulatory systems, including impact assessments and methods for stakeholder participation.
Mujica emphasized that implementation must ultimately be verifiable. Organizations adopting standards or complying with regulations need confidence that others are making comparable efforts and achieving similar outcomes.
Through conformity assessment, he explained, ISO and IEC aim to enable “one standard, one test, one certificate” that can be recognized across markets.
Using and testing the standards that already exist
Philippe Metzger, Secretary-General and CEO of IEC, encouraged governments not to begin from scratch when translating AI policies into practice.
Foundational international standards already exist and can provide reference points for national frameworks. Governments can use them to begin implementing high-level objectives while also contributing to their continued development.
“Don’t start from scratch,” Metzger advised. “If you have high-level policy objectives, leverage the standards that exist.”
Because international standardization is a bottom-up, multi-stakeholder process, governments can help by mobilizing experts from industry, academia, civil society, and other relevant communities to participate. Each country has an opportunity to contribute to the evolution of standards and ensure that national experience informs the international process.
Metzger also stressed the importance of testing standards in real-world settings. Many AI standards are relatively recent, meaning that standards organizations, policymakers, and users still need evidence of how they perform once incorporated into national systems.
He pointed to public procurement as one possible testing ground. The purchase of AI-enabled healthcare systems, for example, immediately raises questions about performance, safety, impact, and assurance. Studying such cases can help determine whether standards are producing their intended economic and social outcomes and where further refinement may be needed.
Why shared challenges require interoperable solutions
Seizo Onoe, Director of ITU’s Telecommunication Standardization Bureau, illustrated the importance of interoperability through the example of deepfakes and misinformation.
Governments increasingly recognize the need to act against manipulated content, but many technical solutions and approaches already coexist.
“If each country adopts different solutions, the effect is very limited. We need a common solution,” Onoe said.
International standards can help create common solutions and ensure that tools developed by different organizations and industries can work together. This requires coordination not only among standards bodies, but also with companies and industry initiatives developing their own methods for content authentication and verification.
Onoe highlighted existing collaboration on AI and multimedia authenticity as an example of how standards organizations can help governments turn policy priorities into practical measures.
The example demonstrates the wider role of interoperability in AI governance. Standards do not require every country or organization to use identical technologies. Instead, they provide the common foundations that allow different approaches to remain compatible and effective across borders.
From standards takers to standards makers
The discussion then turned to inclusion and the participation of developing countries and smaller states in shaping AI standards.
Onoe described global outreach as a central priority for ITU. Through training, fellowships, implementation support, and initiatives including the AI Skills Coalition, ITU aims to strengthen the capacity of developing countries to participate in standardization and adopt standards in practice.
Metzger argued that AI tools themselves could make standards more accessible. By helping users navigate and understand technical content, AI may enable more countries and organizations to apply standards in their daily work and contribute informed feedback to their development.
However, access to documents is not enough. Meaningful participation also requires training, technical workshops, knowledge exchanges, and investment in the infrastructure needed to develop and deploy AI.
Metzger emphasized the importance of electricity and data centers in this context. Without the underlying electrotechnical standards that support digital infrastructure, AI systems cannot operate reliably or sustainably. As countries expand their data-center capacity, standards can contribute to energy efficiency and more sustainable AI development.
Mujica described diversity and inclusion as central to the ISO system. Nearly half of the approximately 80 countries participating in the joint ISO and IEC committee on AI are developing countries, but representation alone does not guarantee influence.
He identified several dimensions of inclusion. Countries need the capacity to send the right experts to international discussions. Standards development must also involve more than AI specialists, incorporating the perspectives of governments, consumers, industry, and the communities affected by the technology.
National committees play an important role by bringing these stakeholders together and ensuring that experts participating internationally represent a broader national position.
Implementation is the final dimension. Mujica cautioned that publishing a standard does not mark the end of the process, but the beginning. Users need support to understand, apply, and evaluate standards in their own contexts.
The invisible infrastructure of AI
Asked to identify what is most important in moving standards from principles to practice, the speakers highlighted three closely connected priorities: capacity building, investment, and collaboration.
Mujica emphasized the knowledge and institutional capacity required to participate in and apply standards. Metzger pointed to investment in both physical infrastructure and standards expertise. Onoe selected collaboration, reflecting the need for standards organizations, governments, industry, and other stakeholders to work together.
Metzger described standards as a form of “invisible infrastructure.” They may not be apparent to the people using an AI-enabled service, but they help ensure that technologies are compatible, reliable, and capable of operating across markets and ecosystems.
Turning AI principles into practice will therefore require more than common aspirations. It will depend on shared language, tested processes, interoperable systems, measurable assurance, and the meaningful participation of countries and stakeholders in shaping how those systems work.
