国际商会国际仲裁院-AI治理标准

July2025|AI governance and standards|1Policy paper Executive summaryThe paper responds to the growing risk of fragmentation in global governance of artificialintelligence (AI). As countries and regions develop their own AI laws, policies and regulations,and even standards, divergent approaches are emerging. The paper highlights the important roleof international and market-driven standards in supporting a broad range of governmentapproaches to AI oversight and discusses how standards can help bridge legal differencesacross jurisdictions.Standards can provide consistent, practical solutions and guidance to comply with laws, policiesand regulation. When governments reference standards as the means to implement non-technicalor high-level performance requirements, they avoid writing technical requirements that mayintroduce unnecessary costs to make products/services available in their jurisdiction. National orregional technical requirements introduced through regulation or standards create complexity forbusinesses of all sizes, increase compliance costs, limit national productivity gains by impeding AIadoption and risk slowing cross-border collaboration and innovation.To ensure effective and interoperable AI governance, greater adoption of standards is essential.Businesses and governments can also bring critical expertise and operational insight to thestandards development process itself. Promoting the use of market-driven standards can reduceduplication, improve regulatory coherence and support policy objectives.1.Why we need standards and what they help us achieveFrom companies developing algorithms to those deploying AI services or systems for their end-users, each participant in the AI supply chain needs clear, consistent guidelines. Internationalstandards are a vital tool for establishing these shared expectations. They serve as a foundationfor fostering interoperability, providing the means for regulatory alignment and facilitating theglobal dissemination of AI innovation.Standards bodies maintain the standards they produce and regularly determine if each standardshould be revised, confirmed or withdrawn. This ability to evolve alongside AI technologies is animportant feature of standards (in comparison to other mechanisms) that keep them relevantand effective over time.What are standards?Standards are documents specifying requirements, guidelines, or characteristics of a product,service, process, or system.They are developed in rules-based, voluntary, multistakeholder organisations that can behorizontal or sectoral and can be national, regional or international. Examples of prominentinformation technology international and market-driven standards bodies include the InternationalOrganisation for Standardisation (ISO), the International Electrotechnical Commission (IEC), theInternational Telecommunications Union (ITU), the Institute of Electrical and Electronics Engineers(IEEE), the Internet Engineering Task Force (IETF) and the European Telecommunications StandardsInstitute (ETSI).Standards development organisations tend to have mature governance systems, particularly dueto process requirements1that take the time needed to elicit contributions from a broad set of1World Trade Organisation (WTO), Principles for the Development of International Standards, Guides andRecommendations (2024). July2025|AI governance and standards|2 July2025|AI governance and standards|3stakeholders and establish consensus among them. The process is intended to produce high-quality results that reflect the best available technical solutions or guidance to build public trustand legitimacy especially in sectors like health, justice and education. Regulation can benefit fromthis deliberative process by citing standards to frame a regulatory approach or set more specificrequirements to carry out the regulation.How are standards and other soft law mechanisms used in relation toAI technologies?There are a variety of soft law mechanisms available that can be used to govern informationtechnology: standards, open-source software (OSS), high-level expert group findings, requirementsmandated through the supply chain (i.e. requirements imposed through procurement), codes ofconduct and guidelines.AI systems and solutions are often built from parts produced by different actors, often in differentjurisdictions. Standards contribute to the vital objective of assuring responsible, safe, secure andinteroperable AI systems and solutions by fostering technical consistency and regulatoryalignment. This harmonisation is essential for maintaining consistency in these practices andcompliance across global markets. In the field of AI, some standards are developed with the goalof harmonising foundational concepts or to promote responsible AI management practices.Standards can also provide the means to address broad principles, including principles that aredefined in laws and regulation. Other standards are being developed to manage safety andsecurity ri