Borderless Trust Through Snap & Verify

Team Members & Divisions

• Dominic Chan - GDP CIO
• Li Lu - GDP DAIP
• Lynette Kwok - SCG P&E
• Poh Quan Wei - GDP CIO
• Tan Cher Shen - GDP CMIP

Problem Statement & Problem Formulation Process

The Challenge

Government agencies worldwide struggle with efficiently verifying the authenticity of foreign-issued documents. The lack of a reliable verification process leads to processing delays and increased fraud risks, ultimately hindering effective public service delivery. While full legalisation provides the highest level of assurance, it is often costly, time-consuming, and impractical. Manual validation, on the other hand, is prone to errors and varies across officers, creating inconsistencies. This leaves a gap between exhaustive verification and blind acceptance, highlighting the need for a more efficient, mid-level verification solution.

Formulating the Problem

Our team conducted extensive research and engaged with members of the International Council for Information Technology in Government Administration (ICA) to understand the pain points experienced by various countries. We identified key problem areas:

• Legalisation of documents is too costly and time-consuming for routine verifications, making large-scale implementation impractical for all use cases.
• In some cases, legalisation of documents is not possible.
• Blind acceptance of foreign documents increases fraud risks.
• Manual verification is inefficient and inconsistent in document inspection and cross-referencing.

Based on these insights, a solution is needed that ensures both efficiency and security. We set out to develop an AI-driven mid-level verification tool designed to assist officers in quickly identifying obvious fraud. This tool aims to reduce unnecessary burdens on both government agencies and citizens, enhancing document verification, strengthening fraud detection, and enabling seamless adoption.

Solution: Borderless Trust

Overview

Borderless Trust is an AI-powered document verification platform that leverages both customised models and LLMs to assess document authenticity. Our solution provides domain-specific analysis tailored to meet particular needs while providing comprehensive insights that deliver a broader perspective on each document.

Key Features

• 🔍 LLM-powered document analysis: Uses GPT-4V to assess document authenticity by analysing holistic cues rather than relying on static pattern matching.
• 🧠 AI-driven vision model & NLP: Extracts, analyses, and cross-checks key document details for accuracy.
• 🎯 Tailored to specific requirements: Trained with real-world forged documents and document specifications to enhance accuracy and practical application.
• 📸 User-friendly interface: Simple to use interface for verification officers to perform document verification—simply snap a photo or upload an image.
• 🔗 API: Designed for easy integration into existing government workflows and seamless adoption by government agencies.
• 🌍 Platform for crowdsourced document ingestion: Issuing authorities can submit details of their official documents into our system to train and improve our document analysis models.

Impact & Outcomes Analysis

Expected Impact and Benefits

⏳ 50% Reduction in Manual Verification Time

Through AI-assisted analysis, the solution significantly accelerates document checks while maintaining high levels of accuracy. This reduction in manual effort frees up valuable time for government agencies, allowing them to focus on more complex tasks and improving overall operational efficiency.

🛑 Lower Fraudulent Document Acceptance Rates

By leveraging advanced AI-driven approaches, the system is capable of detecting subtle inconsistencies and anomalies that may elude human inspectors. This results in a marked reduction in the acceptance of fraudulent documents, enhancing the integrity of government processes.

🚀 Scalable Adoption Potential

Designed for ease of deployment, the solution can be rapidly scaled across multiple countries. This scalability opens up the potential for adoption by government agencies worldwide, helping them streamline document verification in a cost-effective manner.

Pilot Testing & Early Results

We have presented our solution to a selected group of countries, including Botswana, Canada, Israel, and Portugal. Botswana and Canada have further signed up to pilot test our prototype, to demonstrate and evaluate the solution in real-world scenarios. Early feedback indicates promising outcomes. These results validate the solution’s potential and set the stage for broader testing and implementation across other countries.

Future Steps

To enhance Borderless Trust further, we plan to:

• Expand model training with diverse real-world data to improve robustness against evolving document forgeries.
• Integrate additional AI-driven verification layers for higher confidence assessments.
• Partner with more governments to refine the platform based on diverse requirements.
• Develop an API version for easy integration with existing government systems.