Position: Safety and Fairness in Agentic AI Depend on Interaction Topology, Not on Model Scale or Alignment

Abstract

As large language models are increasingly deployed as interacting agents in high-stakes decisions, the AI safety community assumes that safety properties of individual models will compose into safe multi-agent behavior. This position paper argues that this assumption is fundamentally mistaken. In agentic AI, safety is determined by interaction topology, not model weights. When agents deliberate sequentially or aggregate via parallel voting with a judge, the structure of information flow and deci

Categories

@misc{bajaj2026position,
  title = {{Position: Safety and Fairness in Agentic AI Depend on Interaction Topology, Not on Model Scale or Alignment}},
  author = {Tanav Singh Bajaj and Nikhil Singh and Karan Anand and Eishkaran Singh},
  year = {2026},
  month = may,
  eprint = {2605.01147},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2605.01147}
}