Abstract:AI embedded in social governance plays an important role in innovating social governance methods and impro- ving social governance ability. At the same time, the application of AI in social governance has brought risks, such as the se- curity risks of governance technology, the substitution crisis of governance subjects, the lack of value of governance behavior and the ambiguity of governance responsibilities. The technical complexity and uncertainty of AI, the dual effects of enabling governance and infringement of rights, and the reflexivity of both governance tools and governance objects are its risk sources. The risk avoidance of AI embedded in social governance is a “system engineering”, which should be committed to building a comprehensive public policy framework. In short, we should strengthen the root cause management of AI risks based on data and algorithm management; optimize and improve social policies as a guarantee to highlight the value rationality of technological governance behavior; take risk assessment and social experiment as means to carry out AI risk monitoring and early warning; take social cooperation and co - governance as the frame-work to promote the collaborative governance of AI risks.