cs.AIFeb 12, 2026

GT-HarmBench: Benchmarking AI Safety Risks Through the Lens of Game Theory

Authors: Pepijn Cobben, Xuanqiang Angelo Huang, Thao Amelia Pham, Isabel Dahlgren, Bernhard Schölkopf, Terry Jingchen Zhang, Zhijing Jin

Organizations: ETH Zürich · Berea College · Vector Institute · University of Toronto · 

Abstract

Frontier AI systems are increasingly capable and deployed in high-stakes multi-agent environments. However, existing AI safety benchmarks largely evaluate single agents, leaving multi-agent risks such as coordination failure and conflict poorly understood. We introduce GT-HarmBench, a benchmark of 1,535 high-stakes scenarios spanning game-theoretic structures such as the Prisoner's Dilemma, Stag Hunt and Chicken. Scenarios are drawn from realistic AI risk contexts in the MIT AI Risk Repository. Across 15 frontier models, agents fail to choose socially beneficial actions in 39% of high-stakes cases, such as military escalation, election manipulation, and medical malpractice. We measure sensitivity to game-theoretic prompt framing and ordering, and analyze reasoning patterns driving failures. We further show that game-theoretic interventions improve socially beneficial outcomes by up to 31%. Our results highlight substantial reliability gaps and provide a broad standardized testbed for studying alignment in multi-agent environments. The benchmark and code are available at https://github.com/Flecart/gt-harmbench.

Figures & tables

Appendix figures & tables18 assets

Supplementary material from the paper’s appendix.

Appendix

Explore similar work

CardsList
  1. AM-Bench: A Unified Taxonomy and Evaluation Suite for Agentic Misalignment

    Oct 16, 2026Eric Zhang, Terry Jingchen Zhang, Chijioke Ugwuanyi +3AI Agent Safety BenchmarksLLM Agent Safety

  2. GuardianAgentBench: Where Agents Fail and How to Guard Them

    Jul 23, 2026Vishal Ishwar Naik, Chenyu Xu, Donna Dong +5AI Agent ReliabilityLLM Guardrails