KidnapRAG: A Black-Box Attack for Hijacking Reasoning in Agentic Retrieval-Augmented Generation Systems Preprint

Choi, C., Kim, E., Lee, K., Chun, Y., Jeong, J., Kim, E., Oh, M., Jang, J., & Chang, B. (2026).

Industry-academia collaboration with Prof. Buru Chang's group (Korea University, lead author). I led the KT side of the project, am fourth author on the paper, and hold first-inventor priority on the registered Korean patent (Jul 2026). ACL ARR 2026 May submission.

arXiv:2607.00422 · Code

ARS-Bench: Benchmarking Security Risks Across the Agentic RAG Attack Lifecycle Under review

Choi, C., Kim, E., Lee, K., Chun, Y., Jeong, J., Kim, E., Oh, M., Jang, J., & Chang, B. (2026). ACL Rolling Review, August 2026.

A benchmark that evaluates security risks across the entire agentic RAG attack lifecycle, from adversarial document ingestion to final response generation, structured along five security-critical dimensions: document ingestion, retrieval, reasoning, final response, and efficiency. Tracing attack effects through the lifecycle enables fine-grained diagnosis of where vulnerabilities enter and how defences should be targeted.

Paper

Silent Failure of Moral Alignment in Large Language-Model Evaluators Under review

Chun, Y. Under review at Scientific Reports (2026).

Reproducibility release

Moral Orientation and Calibration: Coupled in Human Annotators, Separable in Judge LLMs Accepted

Chun, Y. (2026). ICML 2026 Pluralistic Alignment Workshop.

Paper · Code

A Hybrid Corpus-to-Taxonomy Pipeline for Physical AI Risk Cards

Chun, Y. (2026). Technical Report v1.0, Responsible AI Department, Frontier AI Lab, KT Corporation.

Documents the data collection, deduplication, and hierarchical assignment pipeline behind the Physical AI Risk Taxonomy, with reproducibility artefacts.

AI Technology Specialization and National Competitiveness

Chun, Y., Hur, J., & Hwang, J. (2024). PLOS ONE, 19(4), e0301091.

DOI

The Nexus of Artificial Intelligence and Green Innovation: A Cross-Density Analysis at the Country Level

Chun, Y., & Hwang, J. (2024). Journal of the Knowledge Economy, 16(1), 1688–1716.

DOI

The Moderating Role of Design Innovation in the Relationship Between Technology Complexity and Firm Performance

Hur, J., Chun, Y., Hwang, J., & Kim, K. (2025). Technology Analysis & Strategic Management, 37(11), 1980–1993.

DOI

When Fear Becomes Beauty: Genre as Context in Human and AI Aesthetic Judgment

Wang, I. W., & Chun, Y. (2026).

Cross-institutional collaboration between the Singapore AI Safety Hub (SASH) and KT Frontier AI Lab.

Method and Apparatus for Automatically Constructing a Physical AI Risk Taxonomy Filed 2026

Chun, Y. et al. (2026) [P20260372].

Korean patent application filed 3 August 2026, covering the automated construction of a risk classification system for Physical AI, arising from the Physical AI Risk Taxonomy project.

Method, Server and Computer Program for Dynamically Adjusting Value Information Filed 2025

Chun, Y., Park, Y., Yoon, J., & Lee, S. (2025) [P20250217].

A value-alignment method for generative AI combining multi-agent debate with reinforcement-learning-based reward modelling, so systems can track evolving ethical and social values without continuous human intervention. Received the Excellent Patent Award, KT R&D Center Invention Competition 2025.

Reasoning-Chain Hijacking Attack and Detection for Agentic AI Registered 2026

Chun, Y. (first inventor), Choi, C., Kim, E., Lee, K., Jeong, J., Kim, E., Oh, M., Jang, J., & Chang, B. (2026).

KT-funded and filed ahead of the corresponding paper, with priority held as first inventor.