$PC^2$: Politically Controversial Content Generation via Jailbreaking Attacks on GPT-based Text-to-Image Models

TL;DR

This paper introduces $PC^2$, a novel black-box framework that bypasses safety filters in GPT-based text-to-image models to generate politically sensitive content, revealing significant security vulnerabilities.

Contribution

It presents the first political jailbreaking method exploiting linguistic vulnerabilities in T2I safety filters, with a new benchmark of politically sensitive prompts.

Findings

Achieves up to 86% success rate in bypassing safety filters

Reveals vulnerabilities in current T2I safety mechanisms

Provides a benchmark of politically sensitive prompts

Abstract

The rapid evolution of text-to-image (T2I) models has enabled high-fidelity visual synthesis on a global scale. However, these advancements have introduced significant security risks, particularly regarding the generation of harmful content. Politically harmful content, such as fabricated depictions of public figures, poses severe threats when weaponized for fake news or propaganda. Despite its criticality, the robustness of current T2I safety filters against such politically motivated adversarial prompting remains underexplored. In response, we propose $PC^2$, the first black-box political jailbreaking framework for T2I models. It exploits a novel vulnerability where safety filters evaluate political sensitivity based on linguistic context. $PC^2$ operates through: (1) Identity-Preserving Descriptive Mapping to obfuscate sensitive keywords into neutral descriptions, and (2) Geopolitically Distal Translation to map these descriptions into fragmented, low-sensitivity languages. This strategy prevents filters from constructing toxic relationships between political entities within prompts, effectively bypassing detection. We construct a benchmark of 240 politically sensitive prompts involving 36 public figures. Evaluation on commercial T2I models, specifically GPT-series, shows that while all original prompts are blocked, $PC^2$ achieves attack success rates of up to 86%.