Skip to content
Search
paperSeptember 2026Unreviewed

Data poisoning in LLM multiagent societies: Social proof drives collective decision failure in financial deliberation

Jeongsu Park, Yuji Lim, Geonwoo Kim, Taehyeon Yun, Moohong Min

View resource

Abstract

Abstract Large language model (LLM) multiagent systems are increasingly deployed for high‐stakes financial deliberation, but individually secure LLMs may become vulnerable when embedded in a peer‐to‐peer agent society. We study data poisoning in such societies using Moltbook, a deliberation test bed modeled after the agent‐only social platform. A single malicious agent representing just 14% of a seven‐agent society drives collective accuracy fro

Categories

Data Poisoning

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{park2026data,
  title = {{Data poisoning in LLM multiagent societies: Social proof drives collective decision failure in financial deliberation}},
  author = {Jeongsu Park and Yuji Lim and Geonwoo Kim and Taehyeon Yun and Moohong Min},
  year = {2026},
  month = sep,
  doi = {10.4218/etrij.2026-0186},
  url = {https://doi.org/10.4218/etrij.2026-0186}
}