Information Suppression in Large Language Models: Auditing, Quantifying, and Characterizing Censorship in DeepSeek

TL;DR

This paper investigates how DeepSeek, a Chinese open-source LLM, suppresses sensitive information through an auditing framework, revealing internal reasoning suppression and potential propaganda amplification.

Contribution

It introduces a novel auditing method to detect semantic-level information suppression and characterizes censorship behaviors in DeepSeek.

Findings

DeepSeek suppresses references to transparency and civic issues.

Sensitive content appears in internal reasoning but is omitted in outputs.

The model sometimes amplifies state propaganda language.

Abstract

This study examines information suppression mechanisms in DeepSeek, an open-source large language model (LLM) developed in China. We propose an auditing framework and use it to analyze the model's responses to 646 politically sensitive prompts by comparing its final output with intermediate chain-of-thought (CoT) reasoning. Our audit unveils evidence of semantic-level information suppression in DeepSeek: sensitive content often appears within the model's internal reasoning but is omitted or rephrased in the final output. Specifically, DeepSeek suppresses references to transparency, government accountability, and civic mobilization, while occasionally amplifying language aligned with state propaganda. This study underscores the need for systematic auditing of alignment, content moderation, information suppression, and censorship practices implemented into widely-adopted AI models, to ensure transparency, accountability, and equitable access to unbiased information obtained by means of these systems.