Algorithms, as an important way of connecting people, information, and resources, are being integrated into the intelligent process of libraries in China, reshaping the structural relationship and interactive rules among them. However, the "non-neutrality" of algorithms brings challenges to the data processing and service of libraries, so it has great significance to examine and sort out negative effects of algorithms during information service processes of smart libraries. Firstly, this article introduces critical algorithm theory and takes it as an analytical framework to explore risks of smart libraries from three aspects: information bias, cultural masking, and the lack of subjectivity. Subsequently, internal reasons for the formation of these risks are disclosed, including algorithmic biases, value biases in algorithms, and covert algorithmic control. Finally, risk defense suggestions which are algorithm co-construction, cultural co-creation, and institutional consensus, are provided to achieve the goal of providing effective information services and enhancing cultural identity.