MaMa: A Game-Theoretic Approach for Designing Safe Agentic Systems

Fuente: arXiv
Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Nöther, Jonathan, Singla, Adish, Radanovic, Goran
Format: Preprint
Veröffentlicht: 2026
Schlagworte:
Online-Zugang:
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866913156116250624
author Nöther, Jonathan
Singla, Adish
Radanovic, Goran
author_facet Nöther, Jonathan
Singla, Adish
Radanovic, Goran
contents LLM-based multi-agent systems have demonstrated impressive capabilities, but they also introduce significant safety risks when individual agents fail or behave adversarially. In this work, we study the automated design of agentic systems that remain safe even when a subset of agents is compromised. Inspired by Stackelberg security games, we formalize this problem as a game between a system designer (the Meta-Agent) and a best-responding Meta-Adversary that selects and compromises a subset of agents to minimize safety. We propose Meta-Adversary-Meta-Agent (MaMa), a novel algorithm inspired by this formalization for automatically designing safe agentic systems. Our approach uses LLM-based adversarial search, where the Meta-Agent iteratively proposes system designs and receives feedback based on the strongest attacks discovered by the Meta-Adversary. Empirical evaluations across diverse environments show that systems designed with MaMa consistently defend against worst-case attacks while maintaining performance comparable to systems optimized solely for task success. Moreover, the resulting systems generalize to stronger adversaries, as well as ones with different attack objectives or underlying LLMs, demonstrating robust safety beyond the training setting.
format Preprint
id arxiv_https___arxiv_org_abs_2602_04431
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle MaMa: A Game-Theoretic Approach for Designing Safe Agentic Systems
Nöther, Jonathan
Singla, Adish
Radanovic, Goran
Machine Learning
Computer Science and Game Theory
LLM-based multi-agent systems have demonstrated impressive capabilities, but they also introduce significant safety risks when individual agents fail or behave adversarially. In this work, we study the automated design of agentic systems that remain safe even when a subset of agents is compromised. Inspired by Stackelberg security games, we formalize this problem as a game between a system designer (the Meta-Agent) and a best-responding Meta-Adversary that selects and compromises a subset of agents to minimize safety. We propose Meta-Adversary-Meta-Agent (MaMa), a novel algorithm inspired by this formalization for automatically designing safe agentic systems. Our approach uses LLM-based adversarial search, where the Meta-Agent iteratively proposes system designs and receives feedback based on the strongest attacks discovered by the Meta-Adversary. Empirical evaluations across diverse environments show that systems designed with MaMa consistently defend against worst-case attacks while maintaining performance comparable to systems optimized solely for task success. Moreover, the resulting systems generalize to stronger adversaries, as well as ones with different attack objectives or underlying LLMs, demonstrating robust safety beyond the training setting.
title MaMa: A Game-Theoretic Approach for Designing Safe Agentic Systems
topic Machine Learning
Computer Science and Game Theory
url https://arxiv.org/abs/2602.04431