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

arXiv: https://arxiv.org/abs/2602.12316 · HTML: https://mlsanity.com/paper/2602.12316/

## Key facts

- Cited by 5 papers in the map (data through 14 Oct 2026).
- Builds on 5 of its 53 references that are in the map.
- First posted on arXiv 12 Feb 2026.

## Builds on (most cited first, 5 of 5)

- [Cooperate or Collapse: Emergence of Sustainable Cooperation in a Society of LLM Agents](https://mlsanity.com/paper/2404.16698/) (2024)
- [GTBench: Uncovering the Strategic Reasoning Limitations of LLMs via Game-Theoretic Evaluations](https://mlsanity.com/paper/2402.12348/) (2024)
- [Put Your Money Where Your Mouth Is: Evaluating Strategic Planning and Execution of LLM Agents in an Auction Arena](https://mlsanity.com/paper/2310.05746/) (2023)
- [From Natural Language to Extensive-Form Game Representations](https://mlsanity.com/paper/2501.17282/) (2025)
- [Can LLMs effectively provide game-theoretic-based scenarios for cybersecurity?](https://mlsanity.com/paper/2508.05670/) (2025)

## Built on by (most cited first, 5 of 5)

- [CoopEval: Benchmarking Cooperation-Sustaining Mechanisms and LLM Agents in Social Dilemmas](https://mlsanity.com/paper/2604.15267/) (2026)
- [Evaluating Cooperation in LLM Social Groups through Elected Leadership](https://mlsanity.com/paper/2604.11721/) (2026)
- [WARBENCH: A Comprehensive Benchmark for Evaluating LLMs in Military Decision-Making](https://mlsanity.com/paper/2603.21280/) (2026)
- [Strategic Heterogeneous Multi-Agent Architecture for Cost-Effective Code Vulnerability Detection](https://mlsanity.com/paper/2604.21282/) (2026)
- [When Agents Lie: Premeditation, Persistence, and Exploitation in Repeated Games](https://mlsanity.com/paper/2607.05132/) (2026)

## Abstract (arXiv)

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. The benchmark and code are available at https://github.com/causalNLP/gt-harmbench.