Focusing on the issues of how the US intelligence ethics code evolves, how it stages and why it turns, this paper studies a multi-source corpus composed of legal and administrative documents, congressional reports, judicial precedents, decldeclfied archives, academic papers and think tank reports based on the framework of historical institutionalism and the method of computational text analysis. Through text preprocessing, TF-IDF feature representation, LDA topic modeling, time-series measurement of topic strength and turning point detection, eight core topics are identified, including national security, civil liberties, supervision mechanism, and technical ethics. The evolution of these core topics is divided into four stages: the foundation of Cold War authorization, the reconstruction of Watergate supervision, the swing back of counter-terrorism security, and the rebalance in the digital age. The results show that the intelligence ethics of the United States is not a linear progress, but a continuous adjustment among security, freedom, supervision and technology under the joint effect of external security pressure, domestic political crisis and technological change, showing the overall path characteristics of “authorization expansion – supervision modification – back swing – dynamic rebalance”. The research provides a computable, staged and interpretable analysis path for the research of information ethics, national security governance and technology governance.
