Stopping Agent Smith: Mitigating Recursive Hacking Through Inference Shaping

An AI attacker is software and can be recursively instantiated on compromised machines. This paper formalizes recursive hacking as a threat class, recognizes inference at the target's cost as the controlling resource, and introduces sparsification of inference capability, fail-closed economic defenses, and crowdsourced defense frameworks as the core containment primitives in a novel strategy we term inference shaping. We model the impact of these measures epidemiologically using a SIRVS model defined over ER and scale-free graphs and provide defensive and policy recommendations to mitigate the impact of this emerging threat.