Financial Fragility in Societies of LLM Agents: Coordination Failures and Stabilizing Mechanisms

2026-09-25

AI Summary

This study addresses the systemic fragility arising in LLM-based agents during financial decision-making, where individual self-preservation behaviors precipitate collective coordination failures. To investigate this phenomenon, we propose the FRAIL framework to simulate scenarios such as bank runs and systematically evaluate the intervention efficacy of three interaction mechanisms: compensatory commitments, centralized protocols, and participant coalitions. Our findings reveal that collective failures remain prevalent even in the absence of malicious instructions. Furthermore, we identify early, widespread commitment as a critical temporal pattern for disrupting the self-reinforcing dynamics of defensive behaviors. We also demonstrate that no single mechanism exhibits universal applicability across diverse contexts. Consequently, this work underscores the necessity of system-level evaluation and robustness-oriented design in multi-agent financial systems to mitigate emergent vulnerabilities driven by decentralized agent interactions.

Abstract

Individually protective decisions can produce avoidable collective failures. As large language model (LLM) agents take on greater roles in financial decision-making, financial AI safety must therefore be considered not only at the level of individual agents, but also at the level of the systems they jointly create. We study this problem with FRAIL, a controlled experimental framework that places LLM agents in three dynamic financial environments---bank runs, debt rollover, and reward crowdfunding---where agents' decisions reshape the financial conditions faced by others. Across seven leading LLMs, we find widespread collective fragility even when no agent is instructed to destabilize the system: 77\% of baseline bank-run episodes and 83\% of debt-rollover episodes end in failure. We then compare three interaction mechanisms based on compensated commitments, centralized commitment agreements, and participant-led coalitions. All three improve aggregate outcomes, but no single mechanism performs best across all financial structures. Across mechanisms, successful stabilization shares a common temporal pattern: broad commitment forms early, before defensive behavior becomes self-reinforcing. Our findings show that individually capable agents do not automatically form safe financial systems, highlighting system-level evaluation and interaction design as central problems for financial AI safety. Code is available at https://anonymous.4open.science/r/FinFrail-CF26.

Problem

• Financial Fragility
• LLM Agents
• Coordination Failures
• AI Safety
• Collective Behavior

Innovation

• LLM agents
• financial fragility
• coordination failure
• interaction mechanisms
• AI safety