Red-Teaming the Stable Diffusion Safety Filter
This work highlights critical safety vulnerabilities in widely used AI image generation models, which is a problem for developers and users seeking to prevent misuse.
The researchers demonstrated that the safety filter in Stable Diffusion can be easily bypassed to generate disturbing content, and they reverse-engineered it to reveal it only blocks sexual content while ignoring violence and gore.
Stable Diffusion is a recent open-source image generation model comparable to proprietary models such as DALLE, Imagen, or Parti. Stable Diffusion comes with a safety filter that aims to prevent generating explicit images. Unfortunately, the filter is obfuscated and poorly documented. This makes it hard for users to prevent misuse in their applications, and to understand the filter's limitations and improve it. We first show that it is easy to generate disturbing content that bypasses the safety filter. We then reverse-engineer the filter and find that while it aims to prevent sexual content, it ignores violence, gore, and other similarly disturbing content. Based on our analysis, we argue safety measures in future model releases should strive to be fully open and properly documented to stimulate security contributions from the community.