This article examines how Chinese practices of security governmentality are enacted in everyday online censorship and surveillance/dataveillance of word flows in the Chinese internet. Our analysis of crowdsourced lists of filtered words on the Sina Weibo microblog shows that search engine filtering is based on a two-layer system where short-lived political incidents tend to be filtered for brief periods of time, while words that are conducive to building oppositional awareness tend to be censored more continuously. This indicates a distinction between ‘bad’ and ‘dangerous’ circulations of information from the viewpoint of Chinese internet censorship. Our findings also point out, perhaps counterintuitively, that the ruling Chinese Communist Party is much more inclined to filter words associated with itself than the opposition, or protests, which are usually regarded as the foci of Chinese internet censorship efforts. Our explanation for this is that through surveillance and censorship, the post-totalitarian party-state protects its political hard core against dangerous circulation by trying to prevent public discourse on its leaders and key opponents from going viral. The Chinese online politics of insecurity makes this feasible in a post-totalitarian political order.