Decoupled safety control improves text to image generation safety
Decoupled Safety Control: A Safety-Control Algorithm for Training-Free Safety Guidance
Computer Vision and Pattern Recognition
Summary
Text-to-image AI models sometimes create unsafe or harmful images. The authors found that existing safety methods mix different safety steps together, which can either fail to block unsafe parts or hurt good image quality. They developed Decoupled Safety Control, a new way to guide these models during generation without retraining them. This approach checks the current image state separately to decide how to intervene, making safety guidance more precise and less disruptive.
What this means in practice
- •For ai product developers: Add safer guidance to text-to-image generation features to reduce harmful content while keeping image quality high without retraining models.
- •For content moderation teams: Improve content filtering of AI-generated images by applying state-dependent control methods that better separate unsafe and safe image parts.
Authors
Huilin Zhou, Ruoxi Cheng, Yuhang Wang, Yuming Liu, Minghao Sun, Ruolong Ma, Lan Zhang
Abstract
Diffusion models achieve strong performance in text-to-image generation, but they can still produce unsafe content. Training-free safety guidance offers a practical inference-time alternative that intervenes during sampling without modifying model weights. However, many existing safeguards still rely on a coupled safety-control chain. In this chain, the same unsafe condition simultaneously specifies what should be avoided, indicates the current unsafe signal, constructs the intervention, and determines when that intervention remains active. This coupling can work when unsafe factors are explicit, but becomes less reliable once harmful and benign semantics are entangled in the current generation state, often leaving residual unsafe content or unnecessarily degrading benign-generation quality or prompt fidelity. We address this problem with Decoupled Safety Control (DSC), a plug-in safety-control algorithm for training-free safety guidance. DSC reformulates training-free safety guidance as an explicit state-dependent control process, so the executed intervention is determined from the current generation state rather than inherited directly from a single unsafe condition. Experiments on standard safety benchmarks show that DSC improves representative latent- and text-space safeguards, achieving stronger unsafe-content suppression while preserving benign-generation quality and maintaining efficient inference.