This paper analyzes the structural transition from inference creep to risk acceleration pipelines in AI-assisted software engineering. Inference creep describes the phenomenon where decision scope expands beyond the original intent during AI-assisted development. As such expansions accumulate, they form cascaded decision chains that amplify systemic risk. This paper proposes the concept of risk acceleration pipelines to explain how localized inference creep evolves into systemic governance failures across the software lifecycle. The analysis contributes to the broader Engineering Decision Behavior Governance (EDBG) paradigm by identifying structural drivers of decision amplification in AI-assisted engineering environments.