Abstract: "Bottom line for safety" is one of the core attributes of meaningful innovation, emphasizing the use of bottom line thinking to prevent and resolve major technological security risks. Based on China's reality, this research makes a predictive analysis of the difficulties and key points of future science and technology security governance, demonstrates the connotation and necessity of bottom line thinking, and puts forward a Bottom Line Thinking-Based science and technology security governance system, which includes five core elements of "Party's leadership", "national system", "mass line", "openness and inclusiveness" and "bottom line thinking". Additionally, by means of target analysis, a "goal-demand-system" three-dimensional research and analysis framework for the new security governance system is constructed and Bottom Lines of AI Governance are put forward. Finally, on the basis of theoretical explanation, the paper makes a forward-looking practical research on the design of the Bottom Line Thinking-Based science and technology security governance system architecture through four design cases, i.e. Applied AI management system architecture, General AI management system architecture, IOT system architecture and Public Security management system architecture, which provides a reference for practically developing and applying the Bottom Line Thinking-Based science and technology security governance system for the management department.