Position: Safety and Fairness in Agentic AI Depend on Interaction Topology, Not on Model Scale or Alignment
Tanav Singh Bajaj, Nikhil Singh, Karan Anand, Eishkaran Singh
Why It Matters
What makes this one worth your time
Understanding the role of interaction topology in AI systems can lead to better safety and fairness outcomes, which is crucial for deploying AI in high-stakes environments.
Interaction topology, not model scale or alignment, is crucial for safety and fairness in agentic AI.
Summary
The paper argues that in agentic AI systems, safety and fairness are more influenced by the interaction topology of agents rather than the scale or alignment of individual models. It identifies three topology-driven pathologies—ordering instability, information cascades, and functional collapse—and suggests that these issues are exacerbated by scaling models. The authors propose that agentic AI should be evaluated as a dynamical system, emphasizing the need for robustness across different interaction architectures.
Key contributions
- Identification of topology-driven pathologies in agentic AI systems.
- Argument for treating agentic AI as a dynamical system for safety evaluation.
Notable insights
- Safety in agentic AI is more dependent on how agents interact rather than their individual capabilities.
- Scaling models can exacerbate topology-driven pathologies, contrary to expectations.
Possible limitations
- Not stated in the abstract
Abstract
arXiv:2605.01147v1 Announce Type: new 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 decision coupling dominates outcomes. Evidence across model families and scales reveals three persistent topology-driven pathologies: ordering instability, where system behavior depends primarily on agent sequence; information cascades, where early judgments propagate regardless of correctness; and functional collapse, where systems satisfy fairness metrics while abandoning meaningful risk discrimination. Contrary to intuition, scaling to more capable models strengthens these effects by increasing consensus formation and reducing the challenge of initial decisions. These failure modes are invisible to model-centric evaluation and alignment procedures. We argue that agentic AI must be treated as a dynamical system rather than a collection of aligned components. Interaction topology must become a primary target of safety evaluation and regulation, with systems required to demonstrate robustness across architectural variations before deployment.