Compliance regulation in cross-border trade exhibits multidimensional characteristics—such as inconsistencies in documentation, fragmentation of compliance information, and difficulty in detecting risk propagation—making it challenging to accurately identify compliance-related matters and risks using solely traditional manual methods or Retrieval-Augmented Generation (RAG) technologies. To address these challenges, this paper proposes an enhanced risk identification method based on RAG. This approach leverages a domain-specific knowledge graph, integrates compliance management throughout the entire cross-border trade lifecycle, and incorporates it into an analytical framework. We construct an ontology model across three dimensions—risk themes, governance metrics, and factual entities—thereby building a cross-border trade compliance knowledge graph comprising 32 nodes and 54 semantic edges, which enables structured representation of rules and evidence. Furthermore, we optimize the conventional RAG workflow by employing techniques such as query rewriting, hybrid retrieval, result ranking, and graph prompting, thereby facilitating the extraction of high-confidence compliance-related items. We employ a two-hop weighting method to quantify the intensity of risk propagation and measure the exposure levels of various compliance-related elements, ultimately generating an interpretable corporate risk profile. The correlation patterns observed in our experimental results indicate the existence of latent compliance risks, particularly those associated with trade financing and anti-money laundering (AML) activities. These findings demonstrate that the proposed method is both feasible and practical, providing a technical solution to enhance compliance in cross-border trade operations.

Zhou Liwen. Knowledge Graph-Augmented RAG Approach: For Compliance Risk Identification and ESG Governance Assessment in Multinational Enterprises. The Frontiers of Society, Science and Technology (2026), Vol. 8, Issue 4: 14-20. https://doi.org/10.25236/FSST.2026.080403.

[1] Pan Quan Tao, Zhang Lei, Liu Bailong, et al. A Graph-Based Traffic Prediction Model—Enhanced via Multi-Source Associative Knowledge Graphs [J]. Journal of Computer Science, 2026,53(8):245–256. DOI: 10.11896/jsjkx.250900014.

[2] Chang L. J., Lai C. M. FactualRAG: A RAG-based framework for fact-checking large language models—leveraging IFCN-certified sources [J]. Journal of Intelligent Information Systems, 2026,64(4):1687–1718. DOI: 10.1007/s10844-026-01041-y.

[3] Wang Ying, Wang Liya. Risk Identification and Countermeasures for Cross-Border Data Asset Trading in China [J]. International Trade, 2026(2):27–37.

[4] Yu Zhi, Li Chunmei. A Tokenization-Based Cross-Border Trusted Data Space: Key Technologies, System Implementation, and Future Prospects [J]. *Journal of Telecommunications Science*, 2026.

[5] Tao Rui, Zhang Sulan, Zhou Huiyuan. An image description generation model based on knowledge graphs and a global contextual attention mechanism [J]. Computer Engineering & Design, 2025, Vol. 46, No.5: 1297–1303.

[6] Abo El-Enen M, Saad S, Nazmy T. A review of retrieval-enhanced generation (RAG) models for healthcare applications [J]. Neural Computing & Applications, 2025,37(33). DOI: 10.1007/s00521-025-11666-9.

[7] Beckhauser W J, Fileto R. EchoRAG: A framework for enhancing language models through graph-RAG and contextual learning [J]. *Machine Learning*, 2025. DOI: 10.1007/s10994-025-06859-1.

[8] Huang Yining. Building a Resilient Regulatory Framework to Address Cyber Threats in Cross-Border Digital Trade: A Perspective Based on Systemic Risk Prevention [J]. *Research on Economic Issues*, 2025(4).

[9] Lu Can, Gao Hui, Yang Jianlin. Research on the Interactive Relationship Between Value and Risk of Cross-Border Data Flows Based on Dynamic Simulation [J]. Journal of China Science and Technology Information, 2025, Vol. 44, No.12:1523–1538.

[10] Mao Jie, Hao Zhenxing, Liu Hongzhong. The Impact of National Security on Risks Associated with Cross-Border Capital Flows: An Analysis Based on Disclosed Capital Flow Data [J]. *World Economic Research*, 2025(8).