Harnessing Data and Intelligence for Collective Advantage 2026

Querying Thailand’s forced labour data in the federated system in the POC FIGURE 4 Query Response Query prompt Language processable data – anonymized Labour inspectionsBorder monitoring dataAudit reportsPrevalence reportsWorker surveysRecruitment datan relevant additional datasetsAI services (agents) LLMaaSParticipant n domainChat prompt RAG data and agentParticipant 2 domain Participant 1 domainGovernment data Supply chain dataRAG dataNGO case data and reports RAG dataFederated collective data Data onboarding within domainDiscover and retrieve federated dataReflect query and plan executionQuery and response based on collective intelligence n = any number of participants or datasets; RAG = retrieval-augmented generation; LLMaaS = large language model as a service Note: This figure illustrates the implementation of the POC. The upper section shows how diverse data sources are coupled with AI services to provide a functional interface for users. The lower section depicts the underlying federated architecture and execution steps that enable that workflow so that the system can generate a collective response without centralizing data and while preserving data sovereignty. Source: World Economic Forum and Hewlett Packard Enterprise Scope and design –Categories of data being considered in the POC: Migration flow records, labour inspection results, prosecution data, government surveys, workplace assessments, supply chain data and anonymized worker grievance information from NGOs, trade unions and hotlines. –Safeguards: All datasets are anonymized and used under strict privacy-preserving protocols; no personally identifiable information is shared or transferred. –Governance: Participating institutions maintain full control of their data while contributing aggregated outputs to a shared analytical layer. The Partnership’s Technical Advisory Group (TAG) provides independent guidance on the design and safeguards of the POC. TAG includes representatives from the Partnership’s technical, business and civil society communities, alongside experts in federated data infrastructure and ethical AI. Ethical assurance is also integral to the POC. To reinforce transparency and responsible innovation, Hewlett Packard Enterprise (HPE) has conducted an independent AI ethics assessment of the Global Data Partnership Against Forced Labour’s design and implementation (see Box 7). Harnessing Data and Intelligence for Collective Advantage: Ending Forced Labour in Global Supply Chains 16