The effective management of artificial intelligence risk is as important as the “head goose effect” driven by its spillover. The governance paths under the competitive perspective of instrumentalism and structuralism is difficult to be the leading logic of governance in the field of artificial intelligence. This paper adopts the research idea of “technology bi-directionality”, constructs a tentative conceptual framework of “technology—institution” co-evolution, and takes deep forgery technology risk governance as a typical case analysis. It is found that artificial intelligence risk governance is a process of iterative development and co-evolution of technological governance and institutional governance. Specifically, technological governance has gradually developed from “passive detection” to “active detection”; institutional governance has gradually changed from “indirect supervision” to “direct supervision”; iterative development of technical system has evolved from “temporary coordination” to “permanent coordination”. Effective governance of artificial intelligence risks requires establishing a flexible thinking logic, introducing diachronic analysis and dialectical thinking into the governance process, and timely and reasonable allocation of technical and institutional tools.