paperAugust 2026Unreviewed

Poisoning Agentic Alpha: Adversarial Vulnerabilities Across Roles and Architectures in Multi-Agent Trading Systems

CheolWon Na, Hao Ni, Lukasz Szpruch, Zhang-Yang Wang, Dhagash Mehta, Saurabh Nagrecha, Alejandro Lopez-Lira, Chanyeol Choi, Yongjae Lee, Jee-Hyong Lee

View resource

Abstract

LLM-based multi-agent trading systems, in which specialized agents collaborate through structured communication to produce trading decisions, are moving rapidly from research prototypes to live deployments that control real assets. The same inter-agent communication that makes them effective also exposes them: a corrupted signal can propagate to the final decision and translate into realized financial loss. Unlike prior attacks that presume privileged access to system internals, we restrict the

Framework mappings

OWASP Top 10 for LLM Applications

LLM04Data and Model Poisoning

MITRE ATLAS

AML.T0020Poison Training Data

Suggested from the entry's categories.

Cite

@misc{na2026poisoning,
  title = {{Poisoning Agentic Alpha: Adversarial Vulnerabilities Across Roles and Architectures in Multi-Agent Trading Systems}},
  author = {CheolWon Na and Hao Ni and Lukasz Szpruch and Zhang-Yang Wang and Dhagash Mehta and Saurabh Nagrecha and Alejandro Lopez-Lira and Chanyeol Choi and Yongjae Lee and Jee-Hyong Lee},
  year = {2026},
  month = aug,
  eprint = {2608.24069},
  archivePrefix = {arXiv},
  url = {https://www.semanticscholar.org/paper/788d33b73f8c3b174b9ba58775547c9439f2f81f}
}