3DEditSafe: Defending 3D Editing Pipelines from Unsafe Generation

TL;DR

This paper introduces 3DEditSafe, a novel safety framework for 3D generative editing pipelines that effectively reduces unsafe content propagation during multi-view scene manipulation.

Contribution

It presents the first safety-regularized 3D editing method combining multiple safety guidance techniques to prevent unsafe content in text-driven 3D scene editing.

Findings

3DEditSafe significantly reduces unsafe semantic alignment in 3D edits.

2D safety guidance alone is insufficient for preventing unsafe 3D generation.

There is a safety-quality tradeoff where stronger safety measures may reduce fidelity or introduce artifacts.

Abstract

Recent advances in 3D generative editing, particularly pipelines based on 3D Gaussian Splatting (3DGS), have achieved high-fidelity, multi-view-consistent scene manipulation from text prompts. However, we find that these pipelines also introduce new safety risks when unsafe prompts produce edits that are propagated and optimized across views. In this work, we study unsafe generation in 3D editing pipelines and show that such behavior can lead to coherent, undesirable Not-Safe-For-Work (NSFW) content in the final 3D representation. To address this, we propose 3DEditSafe, a safety-regularized 3D editing framework that constrains unsafe semantic propagation during optimization. 3DEditSafe combines generation-stage safety guidance with rendered-view 3D safety regularization, safe semantic projection, residue suppression, and mask-aware preservation to steer optimization away from unsafe editing directions. We evaluate our approach on EditSplat scenes using an object-compatible unsafe prompt benchmark and show that 2D safety guidance alone is not consistently sufficient to prevent unsafe 3D edits. 3DEditSafe reduces unsafe semantic alignment and view-level attack success rates, while revealing a safety-quality tradeoff in which stronger unsafe suppression can introduce artifacts or reduce unsafe-prompt fidelity. To our knowledge, this work is the first attempt to study and defend against unsafe generation in text-driven 3D editing pipelines, highlighting the need for safety mechanisms that operate directly on optimized 3D representations.