Is Nepal Geared Up for AI-Powered Risk Management in Customs?

(This article was originally published in the annual publication of Nepal Customs, "Bhansar Smarika," on the occasion of International Customs Day 2024)

1. Introduction:

Every day, customs officials battle an invisible enemy: millions of illicit goods hidden in everything from containers to teddy bears. Intuition alone can’t keep pace with this evolving threat. Enter Artificial Intelligence (AI) - machine learning algorithms now analyze vast datasets, uncover hidden patterns, and pinpoint potential threats with uncanny accuracy. Imagine AI scanning containers in seconds, analyzing declarations, identifying anomalies, and prioritizing inspections - streamlining legitimate trade while tightening the net on illicit activities.

Customs administrations balance trade facilitation, revenue collection, and protecting national security and public health. This requires streamlining clearance, promoting compliance, preventing illicit movement of goods, and safeguarding public health while controlling fraud. However, globalization, technological advancements, and new threats like synthetic documents and trade-based money laundering have added complexity. In this context, “risk” refers to the probability of an event hindering customs enforcement. Risk management defines a structured approach to identify, analyze, mitigate, and monitor potential risks in any customs activity. By embracing this proactive approach, customs strategically allocate resources. High-risk shipments, identified through meticulous analysis, are prioritized for in-depth inspection, while low-risk shipments are expedited, streamlining legitimate trade. Effective risk management empowers customs to enhance effectiveness and efficiency, bolster national security, combat revenue loss, and facilitate faster trade flows, ultimately ensuring the integrity and security of global supply chains.

Unleashing AI’s transformative power in customs risk management holds immense promise. However, unlocking its full potential requires a meticulous strategy. The coming sections dissect the World Customs Organization’s AI initiatives, analyze international best practices, and illuminate both the opportunities and challenges Nepal faces. Our ultimate goal? To forge a roadmap for integrating AI strategically, ensuring it empowers, not replaces, human expertise. This journey aims to empower Nepal’s customs system to seamlessly identify and neutralize illicit activities, paving the way for smoother, more efficient flows of legitimate trade.

2. AI in Risk Management:

Risk management is a fundamental pillar of every successful organization, and customs is no exception. It encompasses a comprehensive process involving all levels, from strategic planning to operational execution. Traditionally, this process consists of:

• Defining the context: Objectives, threats, vulnerabilities, and risk tolerance levels.
• Gathering information and intelligence: Trade declarations, manifest data, intelligence reports, passenger information.
• Risk assessment: Analyzing collected information to identify potential risks and their likelihood and impact.
• Risk evaluation and prioritization: Assessing severity and prioritizing for action.
• Risk treatment: Implementing appropriate mitigation measures like inspections, cargo profiling, or targeted audits.
• Monitoring and review: Monitoring effectiveness and reviewing the overall process.
• Documentation, communication, and consultation: Maintaining clear records and ensuring effective communication.

While human expertise remains vital, AI excels in optimizing the operational level, particularly regarding data analysis and resource allocation. How AI Assists Operational Risk Management:

• Predictive risk assessment and targeting: AI algorithms analyze vast amounts of data, predicting risks like undervaluation, misclassification, and prohibited goods. This enables focusing inspections on high-risk shipments.
• Automated document processing and verifications: AI systems rapidly process and verify documents, identifying inconsistencies and fraud indicators, freeing up officer time and improving clearance speed.
• Trade pattern analysis: AI identifies unusual patterns, indicating potential smuggling attempts, allowing proactive prevention.
• Image recognition and anomaly detection: AI-powered systems analyze baggages and containers, automatically identifying hidden compartments or contraband, enhancing detection capabilities.
• HS classifications: AI assists in accurate HS code classification using Natural Language Processing and large language models, ensuring proper duty assessment and preventing misclassification-related fraud.

Beyond these applications, AI offers groundbreaking advancements, from post-clearance audits to AI-powered supply chain analysis. Passenger clearance can be streamlined through risk-based profiling from Advance Passenger Information (API) and Passenger Name Record (PNR) data, and open-source intelligence gathering can shed light on emerging smuggling trends. However, acknowledging limitations in bias, explainability, and legal frameworks is crucial. While AI is unsuitable for solely dictating decisions, its potential as a powerful assistant for customs officers is undeniable. By responsibly navigating these challenges, we pave the way for a future where AI empowers human expertise to build a more secure and efficient global trade environment.

3. WCO Initiatives and Activities related to AI

As artificial intelligence (AI) reshapes across industries, the World Customs Organization (WCO), an independent intergovernmental body serving 185 member countries, recognizes its immense potential to reshape customs operations. Moving beyond its core activities of rule-setting, international cooperation, and capacity building, the WCO actively fosters AI initiatives to empower its members in leveraging data analytics and intelligent technologies. Since 2019, the WCO has been implementing the BAnd of CUstoms Data Analytics (BACUDA) project to enhance data initiatives focusing on AI implementation. Following are the key initiatives:

• WCO data strategy: Spearheading this movement is the WCO’s data strategy, resting on three pillars: data sharing, building communities of practice, and providing assistance with the transition to data-driven practices. This strategy aims to create a future where customs operations are informed by reliable and shared data, fostering collaboration and innovation across borders.
• Capacity building framework on data analytics: To equip its members with the necessary skills and knowledge, the WCO has established a comprehensive capacity building framework focused on data analytics. This framework promotes evidence-based policymaking, improved data management practices, and increased transparency and trust in customs decision-making. Ultimately, it aims to deliver services aligned with stakeholders’ needs, foster innovation and collaboration, and optimize resource utilization.
• Algorithm Development: On the cutting edge of AI development, the WCO actively promotes the creation of AI algorithms tailored to customs challenges, such as the Dual Attentive Tree-aware Embedding for Customs Fraud Detection (DATE) and AI-HS. These initiatives contribute to the BACUDA data analytics series, a flagship program showcasing the power of AI in tackling complex customs tasks.
• E-learning platform: Furthering its commitment to education, the WCO has established an e-learning platform through CLiKC! offering basics, intermediate, and advanced courses in data analytics and AI. This platform democratizes access to knowledge, ensuring customs officials across the globe have the opportunity to upskill and adapt to the evolving landscape of data-driven customs.
• BACUDA scholarship: Recognizing the transformative potential of AI for developing countries, the WCO established the BACUDA scholarship program. This program selects 12 customs officials from developing countries for five months of in-depth training at Sungkyunkwan University in South Korea, equipping them with the expertise to spearhead AI adoption within their home countries.
• Workshop and Conference: To foster knowledge exchange and collaboration, the WCO organizes regular workshops and conferences like data innovation hubs, national data analytics workshops, regional workshops, and the WCO Technology Conference. These events provide a platform for members to share their experiences, best practices, and challenges related to AI implementation in customs operations.
• Risk Management Compendium: The WCO’s Risk Management Compendium is a dynamic resource that goes beyond operational risk management. Its recent update now explores how AI can be effectively utilized in customs risk management. This invaluable document features real-world use cases and best practices, guiding members toward optimizing their risk management strategies through the power of AI.

In conclusion, the WCO’s embrace of AI signifies a proactive approach to shaping the future of customs. Through its diverse initiatives, the organization empowers its members to leverage data and intelligent technologies, fostering a more efficient, transparent, and secure trade environment for all.

4. Case Studies AI Drives Customs Innovation:

Customs agencies worldwide face a diverse landscape of risk priorities. Revenue protection takes top billing in countries like Nepal, while security dominates concerns in nations like the United States. These varying priorities influence the specific AI solutions implemented for risk management. The following are some countries with their respective considerations

South Korea:

• AI-powered passenger screening: Analyzes passenger information and travel history (API and PNR data) to identify potential risks before arrival.
• Smart surveillance: AI-driven CCTV cameras track individuals at immigration, allowing real-time monitoring and targeted inspections.
• Express cargo inspection: AI X-ray scanners automatically detect contraband in express cargo shipments, improving efficiency and security. Big data risk analysis: KCS Big Data Portal aggregates and analyzes diverse data sources to create a comprehensive risk picture for informed decision-making.

Belgium:

• Combating tariff fraud: Machine learning algorithms identify sudden changes in import profiles after tariff increases, suggesting potential misclassification of origin or commodity codes.
• Entity resolution: Groups similar import entities based on text mining to reconstruct accurate profiles and track potential evasion tactics.
• Prioritizing investigations: Analyzes additional features like economic segment overlap and non-conformity history to prioritize potential fraud cases for further investigation.

Brazil:

• SISAM: AI-powered error detection: Analyzes import declarations using Bayesian Networks to estimate probabilities of errors like incorrect HS codes, missing licenses, and false descriptions.
• Explainable AI: Provides natural language explanations for its predictions, increasing transparency and trust in the system.
• Integrated customs ecosystem: SISAM is connected with other customs systems like Classif (HS code suggestions) and e-Safira (post-clearance feedback) for continuous improvement.

USA

• Image and video analytics: Detecting anomalies, tracking individuals, and recognizing objects of interest in footage from surveillance cameras, drones, and other sources.
• Predictive risk assessment: Analyze passenger and cargo data to predict security threats. Prioritize inspections using multi-modal data for correlation, fusion, and insights.
• Streamlined cargo clearance: Expediting clearance for low-risk shipments through automated analysis of manifests and risk assessments.
• Object identification in streaming video: Capabilities include Counter Small Unmanned Aircraft Systems (CSUAS) for detecting illicit drones and Autonomous Surveillance Towers for maritime domain monitoring.

By tailoring AI solutions to their specific risk priorities, customs agencies can achieve significant improvements in efficiency, security, and revenue collection. This global adoption of AI demonstrates its potential to transform customs operations and safeguard international trade.

5. Opportunities and Challenges for Nepal to Implement AI in Customs

Nepal stands at a crossroads in its customs operations. Embracing Artificial Intelligence (AI) presents immense opportunities to streamline trade, enhance security, and boost revenue collection. However, navigating the challenges that accompany AI adoption is crucial for its successful integration into the Nepali customs landscape.

Opportunities:

• Improved Risk Management: AI-powered risk assessment algorithms can analyze vast amounts of trade data, identifying high-risk shipments for targeted inspections. This prioritizes resources, expedites legitimate trade flow, and combats smuggling and revenue evasion.
• Fraud Detection and Prevention: AI can scrutinize documents, trade patterns, and anomalies to detect fraud attempts like misclassification, undervaluation, and false declarations. This safeguards national security and public health while bolstering revenue collection.
• Enhanced Trade Facilitation: Automated document processing and verification using AI can significantly reduce clearance times for low-risk shipments. This promotes seamless trade flow, attracts foreign investment, and stimulates economic growth.
• Resource Optimization: AI can analyze staffing needs, predict peak periods, and optimize the deployment of personnel. This enhances operational efficiency, reduces redundancy, and ensures optimal resource allocation.
• Leveraging internal resources: Nepal possesses a strong IT team, experienced manpower, WCO pre-accredited risk management experts, and BACUDA scholars – a valuable foundation for building AI expertise and tailored solutions for its customs landscape

Challenges:

• Data Infrastructure and Availability: Implementing AI requires robust data infrastructure and access to high-quality, reliable data. Nepal needs to invest in data storage, management, and sharing practices to support AI initiatives.
• Technology and Expertise: Nepal’s customs officials require training and upskilling in data analytics, AI, and cyber security to effectively operate and manage AI systems. Collaboration with international partners and knowledge-sharing platforms is crucial.
• Ethical Considerations: Algorithmic bias, explainability of decisions, and data privacy concerns need careful consideration. Nepal must develop ethical frameworks and policies to ensure responsible AI implementation in customs.
• Cost and Investment: Initial investments in AI infrastructure, software, and training can be significant. Nepal needs to explore cost-effective solutions, leverage international assistance, and prioritize funding for this transformative technology.
• Legal and Regulatory Frameworks: Clear legal and regulatory frameworks governing data privacy, AI use, and algorithmic decision-making are essential. Nepal needs to adapt its legislation and legal structures to accommodate AI integration in customs.

6. Recommendations for Implementing AI in Nepal Customs:

The potential for AI to transform Nepal’s customs operations is exciting, especially with the nation’s focus on streamlining trade and promoting economic growth. This phased approach outlines a comprehensive roadmap for implementing AI in Nepal Customs:

Phase 1: Laying the Foundation

1. Establish an AI and Innovation Unit: This dedicated unit will be the engine driving AI adoption. It will research, develop, and manage pilot projects, ensuring focused and effective implementation.
2. Build the Infrastructure: Investing in robust hardware and data storage is crucial for running AI applications and maintaining data security.
3. Prioritize Risk Assessment: Develop a pilot project utilizing AI to analyze import declarations, targeting potential undervaluation and misclassification. This initial success story can pave the way for wider adoption.
4. Leverage Open Source Tools: Explore open-source Large Language Models for AI-powered Harmonized System (HS) classification, minimizing resource constraints.

Phase 2: Expanding Applications

5. Prepare for X-ray Analysis: Begin collecting and labeling X-ray image data for future AI-powered analysis. Simultaneously, implement an API Passenger Name Record (PNR) system for enhanced passenger profiling.
6. Gather External Data: Integrate information from other government sources (e.g., intelligence agencies) to enrich risk assessment models.
7. Develop a Risk Dashboard: Create a visual platform showcasing real-time risk insights, enabling efficient decision-making by customs officials.
8. Establish Legal Framework: Develop a robust legal framework guiding the ethical and responsible integration of AI into Nepal Customs operations.

Phase 3: Scaling Up and Refining

9. Strengthen Data Security: Implement robust cybersecurity measures and data privacy policies to ensure the safe and responsible handling of sensitive information.
10. Deploy Risk Targeting Model: Integrate the AI-powered risk assessment model into the live customs system, complementing and optimizing existing targeting methods.
11. Advance X-ray Analysis: Develop an AI model for automated X-ray image analysis, enhancing cargo inspection efficiency and accuracy.
12. Refine Data Visualization: Continuously improve data visualization dashboards to track trade patterns, identify trends, and inform data-driven decision-making.

Phase 4: Continuous Improvement

13. Analyze API PNR Data: Develop a dedicated analytical system for extracting valuable insights from API PNR data, further improving passenger profiling and risk assessment.
14. Automate Document Processing: Implement AI-powered automated document processing to extract data from customs documents, minimizing manual effort and errors.
15. Embrace Constant Learning: Continuously monitor and adapt to the latest advancements in AI technology, ensuring Nepal Customs remains at the forefront of innovation.

Beyond the Phases: Sustaining Progress

• Nurture Collaboration: Encourage continuous experimentation, collaboration, and knowledge sharing within the department to maintain a dynamic and innovative culture.
• Seek External Expertise: Leverage the World Customs Organization’s (WCO) resources and expertise in AI implementation for customs.
• Connect with Peers: Engage in knowledge sharing initiatives with other customs authorities implementing AI, learning from their successes and challenges.
• Invest in Training: Develop and conduct training programs on AI fundamentals, specific applications, and ethical considerations for customs officials.
• Empower Staff Participation: Encourage staff participation in AI workshops, conferences, and exchange programs to foster a deep understanding and buy-in for AI initiatives.
• Retain Talent: Implement policies that attract and retain AI experts, ensuring Nepal Customs has the human capital to drive its AI journey forward.
• Educate the Public: Create awareness campaigns to educate the public about the benefits and applications of AI in customs, building trust and transparency.

By following these recommendations and fostering a culture of innovation, Nepal Customs can harness the power of AI to unlock a new era of efficiency, transparency, and economic prosperity. Let this be the roadmap for Nepal’s journey towards becoming a leading AI-powered customs authority in the region.

References:

World Customs Organization & World Trade Organization. (2022). WCO/WTO Study Report on Disruptive Technologies.

Mikuriya, K., & Cantens, T. (2020). If algorithms dream of Customs, do customs officials dream of algorithms? A manifesto for data mobilisation in Customs. World Customs Journal, 14(2), 3-22.

Kafando, I. (2020). How can Customs better leverage emerging AI technologies for more sustainable and smarter operations? World Customs Journal, 14(2), 143-156.

Belan, M. (2023, Aug 25). AI vs. Humans: Which Performs Certain Skills Better? From https://www.visualcapitalist.com/ai-vs-humans-which-performs-certain-skills-better/

Iordache, E., & Voiculet, A. V. (2007). Customs Risk Management in the European Union. The Romanian Economic Journal.

Wrold Customs organization. (2022). Illicit Trade Report.

World Custom Organization &World Trade Organizaton. (2021). The role of advanced technologies in cross-border trade: A customs perspective.

World Customs Organization. (n.d., [Accessed:2024-01-12]). Risk management compendium. International Organization for Standardization. (2018). isk management - Guidelines and principles (ISO 31000:2018). Geneva, Switzerland.

(n.d.,[Accessed: 2024-01-01]). From world Customs Organization: https://www.wcoomd.org/

(n.a,[Accessed: 2024-01-01]). From WCO BACUDA Project: https://bacuda.wcoomd.org/

Kim, S. a.-C.-T. (2020). DATE (Dual Attentive Tree-aware Embedding for Customs Fraud Detection) . In Proceedings of the 26th ACM SIGKDD Conference on Knowledge Discovery and Data Mining (KDD ’20).

Artificial Intelligence to Harness Key Insights at CBP . (2023, 03 24). From U.S. Customs and Border Protection: https://www.cbp.gov/newsroom/spotlights/artificial-intelligence-harness-key-insights-cbp