Abstract: Artificial intelligence algorithm represented by deep learning, though bearing breakthrough characteristics, is applied in government departments in an embedded mode. Such an application starts from algorithm decision-making, and step by step is found available among administrative rules, administrative power and bureaucratic structure. The bureaucratic structure of government can choose algorithmic decision-making, and the bureaucratic system will not disappear. The influence of algorithm embedded includes instrumentality, embeddedness, autonomy and destructiveness and the controllable algorithm runs out of control when the system is missing or fails. A comprehensive embedding, as for the application of public domain algorithms in China, formed by extensive participation under the high-level promotion, brings forth systematic risk resulted from data leakage and algorithm out of control, a situation which renders an integration of the network security framework a necessity for continuously facilitating the management and control.

Key words: government bureaucracy, algorithmic embedding, algorithmic power, risk response