SoK: Trading Agents or Market Crashers? Dissecting Robustness and Security Failures in Academic Financial LLM Trading Schemes
This paper surveys and stress-tests academic LLM-based financial trading agent designs with FARSIGHT, a framework that evaluates whether they stay sane under turbulent markets and whether they can be compromised through poisoned information sources, direct agent attacks, or malicious agent behavior. The interesting bit is that it treats trading agents as high-stakes, market-coupled systems rather than generic chatbots, showing that robustness bugs and security bugs collapse into the same failure mode—runaway bad trades or even crash-like dynamics—though the evidence is limited to representative academic schemes rather than live production trading systems.
Abstract
Autonomous large language model (LLM) agents are moving rapidly into high-stakes domains, yet existing agentic-AI security studies remain largely domain-agnostic and overlook the distinctive, high-consequence attack surface such settings create. We examine this gap through financial trading agents, a representative case of high-stakes agentic security, where a single compromised agent has direct execution authority over real capital in an adversarial, reflexive market. To this end, we present FARSIGHT (Financial Agent Robustness and Security Investigation and Global Holistic Testing), a framework that performs scheme-level evaluation of financial LLM agents on two axes: robustness under market turbulence (including flash-crash-like scenarios), and security against three attack types: attacks on information sources, attacks on agents, and agent-as-attacker behaviors. Applying FARSIGHT to 15 representative academic schemes, we find that most overlook robustness and realistic adversarial threats: 80% fail at least one core robustness metric and 100% exhibit security vulnerabilities. These two failure modes are inseparable: a small misjudgment can cascade into a market-wide crash on its own, while an adversary can deliberately trigger the same collapse at minimal cost.