Coevolution in generative AI content ecosystems: behaviour-risk feedback, governance chains and local stability boundaries

Abstract Generative artificial intelligence enables producers to reformulate risky content after moderation, creating feedback between behaviour and governance. We examine how joint weakening across governance stages and differences in response speeds affect risk, trust and local stability. We develop a dimensionless, uncalibrated governance--adaptation evolutionary game linking content producers, platform governance units and users. The model includes finite moderation capacity, provenance retention, trust memory and endogenous producer adaptation. In the tested factorial design, jointly weakening the governance channels increased risk and reduced trust and user-experienced quality; under a common favourable reference, joint deterioration exceeded the corresponding positive single-channel losses for all three outcomes. Along a prescribed composite continuation path, spectral analysis indicated that tenfold acceleration of either backlog or provenance dynamics shifted the estimated local-stability boundary toward stronger aggregate feedback, whereas tenfold acceleration of trust-memory dynamics shifted it toward weaker feedback. These results are conditional on the specified structures and operating states: matching the adaptation state alone did not ensure equivalent transient responses or local stability. The framework supports analysis of governance coordination and response timing; application to particular platforms requires empirical calibration and validation.