In many complex systems, constituent agents engage in adversarial behavior without explicit awareness or intent. These "unaware adversaries" operate based on localized objectives and feedback, yet their interactions within a shared environment result in mutual disruption, resource competition, or functional conflict. This paper formally defines this phenomenon and situates it within existing literature on systems theory. We present and analyze several case studies—including closed-loop, open-loop, and designed conflicts—to build a taxonomy of adversarial typologies. By demonstrating how this framework can be used as a generative tool for analysis and design, we derive illustrative mitigation strategies for problems in fields as diverse as generative modeling and internet routing.
Keywords: unaware adversaries, emergent conflict, system dynamics, feedback control loops, multi-agent systems, generative adversarial networks, open-loop systems, BGP, network stability, protocol design
