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

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

Keywords: multi-agent systems, AI safety, alignment, game theory, benchmarking, mechanism design

TL;DR: GT-HarmBench introduces 1,535 multi-agent game-theoretic scenarios in high-stakes AI safety contexts. We find systematic failures in cooperation and frequent defection, but demonstrate that mechanism design interventions can improve outcomes.

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 38% 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 18%. Our results highlight substantial reliability gaps and provide a broad standardized testbed for studying alignment in multi-agent environments.

Email Sharing: We authorize the sharing of all author emails with Program Chairs.

Data Release: We authorize the release of our submission and author names to the public in the event of acceptance.

Paper Type: Standard paper

Submission Number: 69