[2603.20925] Profit is the Red Team: Stress-Testing Agents in Strategic Economic Interactions

Computer Science > Artificial Intelligence arXiv:2603.20925 (cs) [Submitted on 21 Mar 2026] Title:Profit is the Red Team: Stress-Testing Agents in Strategic Economic Interactions Authors:Shouqiao Wang, Marcello Politi, Samuele Marro, Davide Crapis View a PDF of the paper titled Profit is the Red Team: Stress-Testing Agents in Strategic Economic Interactions, by Shouqiao Wang and 3 other authors View PDF HTML (experimental) Abstract:As agentic systems move into real-world deployments, their decisions increasingly depend on external inputs such as retrieved content, tool outputs, and information provided by other actors. When these inputs can be strategically shaped by adversaries, the relevant security risk extends beyond a fixed library of prompt attacks to adaptive strategies that steer agents toward unfavorable outcomes. We propose profit-driven red teaming, a stress-testing protocol that replaces handcrafted attacks with a learned opponent trained to maximize its profit using only scalar outcome feedback. The protocol requires no LLM-as-judge scoring, attack labels, or attack taxonomy, and is designed for structured settings with auditable outcomes. We instantiate it in a lean arena of four canonical economic interactions, which provide a controlled testbed for adaptive exploitability. In controlled experiments, agents that appear strong against static baselines become consistently exploitable under profit-optimized pressure, and the learned opponent discovers probing, ...