Page 52 - AI Standards for Global Impact: From Governance to Action

AI Standards for Global Impact: From Governance to Action

7 Open dialogue on trustworthy AI testing

The Open Dialogue on Trustworthy AI Testing workshop brought together global stakeholders
to address critical gaps and collaboration opportunities in trustworthy AI testing. The workshop
employed an interactive format to facilitate focused discussions across three key pillars of
AI testing collaboration: capacity building, standards and best practices, and institutional
frameworks.

The main objectives of the workshop were to:
1) Identify current gaps in global AI testing capabilities
2) Explore collaboration mechanisms for capacity building in AI testing
3) Discuss standards, best practices, and conformity assessment needs
4) Examine institutional frameworks for international coordination on AI testing

The audience was divided into three thematic groups:

Group 1: Capacity Building

Focus: Understanding global capacity needs and gaps for trustworthy AI testing

Group 2: Standards, Best Practices and Conformity Assessment

Focus: Current best practices, methodologies, standards gaps, and knowledge sharing
mechanisms

Group 3: Institutional Frameworks

Focus: AI governance structures and coordination at the international level

7�1 Outcomes

The outcomes of the discussions of each group are summarised below.

7�1�1 Group 1: Capacity building

The capacity building group identified several critical areas requiring attention:

Current gaps in AI testing capabilities globally

• Terminology around testing remains highly confusing and varies significantly across
different contexts and regions
• Substantial disparities exist between current use-case testing and real-world testing
scenarios
• Limited and unequal access to AI models, which is essential for comprehensive testing

Knowledge areas and institutional capabilities needing development
• Standardized terminologies and metrics for AI testing
• Clear definition of roles for different stakeholders in the testing ecosystem Institutional
capacity, which varies dramatically across different regions, particularly affecting emerging
economies' ability to conduct AI testing for various use cases

40