Corruption as Functional Capture: From Diverted Flow to Self-Reproducing Systemic Environments

Corruption is commonly described through bribery, illegality, moral failure, or abuse of office. These descriptions identify important manifestations but do not capture the broader process pattern shared by institutional, informational, linguistic, organizational, and human–AI failures. This working paper defines corruption functionally as the capture or compromise of a system function such that declared purpose, actual flow, and resulting fruit no longer remain aligned. The formal structure may persist while another regulator redirects money, information, attention, authority, responsibility, or feedback toward a private or parasitic outcome. The paper develops the distinction between healthy function, captured function, and a later “zombie” phase in which the environment reproduces the corrupted pattern without requiring a continuously active initiator. It examines the role of unowned transition gaps, corrupted inputs, false trace, diluted responsibility, products without fruit, and procedures without closed feedback. It further argues that large language models can compress, normalize, and return patterns inherited from a corrupted socio-technical environment, creating a feedback loop in which environmental trace becomes new environmental input. The proposed response is not moral labeling alone but explicit process regulation: preserving clean input, provenance, status, responsibility, transition control, reality feedback, and visible open gaps. AWM is introduced only as an emerging answer to this regulatory need; its technical implementation remains outside the scope of the paper.