The Anomaly Detection for Assistance Delivery (ADAD) initiative

The Anomaly Detection for Assistance Delivery (ADAD) initiative

The Anomaly Detection for Assistance Delivery (ADAD) initiative demonstrates how artificial intelligence can strengthen the integrity and effectiveness of humanitarian assistance at global scale. Developed through a collaboration between the World Food Programme (WFP), CERN, and the Government of Luxembourg, ADAD helps identify potential fraud, duplication, and data quality issues in beneficiary registration systems before assistance is delivered.

WFP supports more than 120 million people worldwide and relies on its SCOPE platform to manage beneficiary data. In many operational contexts, limited access to biometric verification increases the risk of registration errors, duplication, and manipulation. ADAD addresses this challenge using an autoencoder-based neural network that learns what normal registration patterns look like from historical data and automatically flags unusual records for review.

During the demo, participants will see how the system detects a range of anomalies, including demographic inconsistencies, suspicious household structures, geographic contradictions, and other unusual registration patterns. The solution clusters similar anomalies to help data assurance teams focus on investigations where they can have the greatest impact.

Unlike traditional rule-based approaches, ADAD adapts to local contexts and supports human decision-making rather than replacing it. By improving accountability and reducing errors and duplication, ADAD helps ensure that scarce humanitarian resources reach the people who need them most.

Share this speaker
Share This Speaker
  • Organization
  • Profession
Related sessions
9 July 2026
11:00 - 11:15
EST - New York
CST - Beijing
PST - Los Angeles
AWST - Perth, Australia

Are you sure you want to remove this speaker?