Safe T2I Generation

T2I: Text-to-Image

It studies how to keep text-to-image systems from producing harmful images without blocking legitimate requests. Approaches include prompt moderation, safety alignment during training, concept erasure, and inference-time steering or localized repair during denoising. Researchers test these methods on safety benchmarks and red-team attacks, including jailbreaks, unsafe combinations of benign concepts, and image-editing settings, while measuring image quality and prompt fidelity. A central challenge is making defenses robust to changing attacks and context-dependent harms without suppressing benign content, with performance varying across model architectures and generation methods.

Latest papers

53 papers9 papers in the last four weeks, up 200% on the four weeks before. 0.1% of all new papers.

Disciplined Diffusion: Text-to-Image Diffusion Model against NSFW Generation
May 1, 2026Chi Zhang, Changjia Zhu, Xiaowen Li +2Adversarial Robustness Safe T2I Generation
Bellini College of Artificial Intelligence, Cybersecurity and Computing, University of South Florida, Tampa, FL, USA · Department of Electrical and Computer Engineering, College of Engineering, University of South Florida, Tampa, FL, USA
When Safe Concepts Become Unsafe: Multi-Concept Compositional Vulnerabilities in Text-to-Image Models
Apr 17, 2026Chaoshuo Zhang, Yibo Liang, Mengke Tian +7Safe T2I Generation Compositional T2I Generation
School of Cyber Science and Engineering, Xi’an Jiaotong University, Xi’an, China · CISPA Helmholtz Center for Information Security, Saarbrücken, Germany · School of Cyber Science and Engineering, Wuhan University, Wuhan, China
Beyond Text Prompts: Precise Concept Erasure through Text-Image Collaboration
Apr 17, 2026Jun Li, Lizhi Xiong, Ziqiang Li +4Safe T2I Generation Controllable Image Generation
Nanjing University of Information Science and Technology · Southeast University · Nanjing University of Science and Technology
Erasing Thousands of Concepts: Towards Scalable and Practical Concept Erasure for Text-to-Image Diffusion Models
Apr 12, 2026Hoigi Seo, Byung Hyun Lee, Jaehyun Cho +2Safe T2I Generation Diffusion Model Unlearning
Dept. of Electrical and Computer Engineering, Seoul National University, Republic of Korea · IPAI, Seoul National University, Republic of Korea · INMC, Seoul National University, Republic of Korea
EGLOCE: Training-Free Energy-Guided Latent Optimization for Concept Erasure
Apr 10, 2026Junyeong Ahn, Seojin Yoon, Sungyong BaikDiffusion Model Unlearning Safe T2I Generation
KAIST AI · Hanyang University
Reading Between the Pixels: An Inscriptive Jailbreak Attack on Text-to-Image Models
Apr 7, 2026Zonghao Ying, Haowen Dai, Lianyu Hu +5Black-Box Adversarial Attacks Safe T2I Generation
Beihang University Beijing, China · University of Nottingham Ningbo China Ningbo, China · Nanyang Technological University Singapore +2
SafeCtrl: Region-Aware Safety Control for Text-to-Image Diffusion via Detect-Then-Suppress
Apr 5, 2026Lingyun Zhang, Yu Xie, Zhongli Fang +2Safe T2I Generation Adversarial Robustness
School of Computer Science and Technology, Fudan University, China · Institute of Big Data, Fudan University, China · Nanjing University of Information Science and Technology, China
EDGE-Shield: Efficient Denoising-staGE Shield for Violative Content Filtering via Scalable Reference-Based Matching
Apr 4, 2026Takara Taniguchi, Ryohei Shimizu, Duc Minh Vo +3Safe T2I Generation T2I Diffusion Models
SB intuitions
Modular Energy Steering for Safe Text-to-Image Generation with Foundation Models
Apr 2, 2026Yaoteng Tan, Zikui Cai, M. Salman AsifSafe T2I Generation T2I Diffusion Models
University of California Riverside · University of Maryland
Low-Effort Jailbreak Attacks Against Text-to-Image Safety Filters
Apr 2, 2026Ahmed B Mustafa, Zihan Ye, Yang Lu +2Adversarial Attacks on VLMs Safe T2I Generation
School of Computer Science, University of Nottingham, Wollaton Road, NG8 1BB · Department of Intelligent Science, Xi’an Jiaotong-Liverpool University, China, 215123 · School of Informatics, Xiamen University, Xiamen, 361005, China
SafeRoPE: Risk-specific Head-wise Embedding Rotation for Safe Generation in Rectified Flow Transformers
Apr 2, 2026Xiang Yang, Feifei Li, Mi Zhang +3Safe T2I Generation Diffusion Transformer
Fudan University, Shanghai, China · East China University of Science and Technology, Shanghai, China
IDDM: Identity-Decoupled Personalized Diffusion Models with a Tunable Privacy-Utility Trade-off
Apr 1, 2026Linyan Dai, Xinwei Zhang, Haoyang Li +2T2I Diffusion Models Image Generation
The Hong Kong Polytechnic University, Hong Kong, China
DTVI: Dual-Stage Textual and Visual Intervention for Safe Text-to-Image Generation
Mar 23, 2026Binhong Tan, Zhaoxin Wang, Handing WangSafe T2I Generation T2I Diffusion Models
Xidian University Xi’an, China
Not All Latent Spaces Are Flat: Hyperbolic Concept Control
Mar 14, 2026Maria Rosaria Briglia, Simone Facchiano, Paolo Cursi +6Controllable Image Generation Safe T2I Generation
Department of Computer Science, Sapienza University of Rome, Rome, Italy · ItalAI Labs, Rome, Italy · Paradigma Inc., Rome, Italy
IdentityGuard: Context-Aware Restriction and Provenance for Personalized Synthesis
Mar 14, 2026Lingyun Zhang, Yu Xie, Ping ChenAI-Generated Image Watermarking Safe T2I Generation
School of Computer Science and Technology, Fudan University, Shanghai, China · Institute of Big Data, Fudan University, Shanghai, China · Purple Mountain Laboratories, Nanjing, China
OrthoEraser: Coupled-Neuron Orthogonal Projection for Concept Erasure
Mar 12, 2026Chuancheng Shi, Wenhua Wu, Fei Shen +3Safe T2I Generation
University of Sydney Sydney, Australia · National University of Singapore Singapore, Singapore · Adelaide University Adelaide, Australia +1
Prototype-Guided Concept Erasure in Diffusion Models
Mar 9, 2026Yuze Cai, Jiahao Lu, Hongxiang Shi +2Concept Erasure Safe T2I Generation
Fudan University · National University of Singapore
Conditioned Activation Transport for T2I Safety Steering
Mar 3, 2026Maciej Chrabąszcz, Aleksander Szymczyk, Jan Dubiński +3Safe T2I Generation Activation Steering
NASK National Research Institute, Warsaw, Poland · Warsaw University of Technology, Warsaw, Poland · Tooplox +2
Safety-Guided Flow (SGF): A Unified Framework for Negative Guidance in Safe Generation
Mar 2, 2026Mingyu Kim, Young-Heon Kim, Mijung ParkDiffusion Model Guidance Constrained Generative Modeling
Kookmin University · University of British Columbia
When Safety Collides: Resolving Multi-Category Harmful Conflicts in Text-to-Image Diffusion via Adaptive Safety Guidance
Feb 24, 2026Yongli Xiang, Ziming Hong, Zhaoqing Wang +3Safe T2I Generation Diffusion Model Guidance
Sydney AI Centre, The University of Sydney · City University of Hong Kong · TMLR Group, Hong Kong Baptist University
FIGURA: A Modular Prompt Engineering Method for Artistic Figure Photography in Safety-Filtered Text-to-Image Models
Feb 20, 2026Luca CazzanigaPrompt Optimization for T2I Generation T2I Generation
Independent Researcher, Monza, Italy
AEIOU: A Unified Defense Framework against NSFW Prompts in Text-to-Image Models
Dec 24, 2024Yiming Wang, Jiahao Chen, Qingming Li +4Safe T2I Generation Adversarial Robustness
†Zhejiang University · ‡National University of Defense Technology
Rethinking Negative Prompting for Concept Suppression: Unconditional Anchoring and Localized Guidance
Dec 10, 2024Anubhav Jain, Takashi Shibuya, Yuya Kobayashi +4Diffusion Model Guidance Safe T2I Generation
New York University · Sony AI · Sony Group Corporation

May 1, 2026J/K move · Enter open · S save

Disciplined Diffusion: Text-to-Image Diffusion Model against NSFW Generation

Chi Zhang, Changjia Zhu, Xiaowen Li, Yao Liu+1

Bellini College of Artificial Intelligence, Cybersecurity and Computing, University of South Florida, Tampa, FL, USA · Department of Electrical and Computer Engineering, College of Engineering, University of South Florida, Tampa, FL, USA

Adversarial Robustness Safe T2I Generation T2I Diffusion Models

Open paper AskPDF arXiv

Text-to-image (T2I) diffusion models have the ability to build high-quality pictures from text prompts, but they pose safety concerns because they can generate offensive or disturbing imagery when provided with harmful inputs. Existing safety filters typically rely on text-based classifiers or image-based checkers that completely block the output upon detecting a threat, issuing an explicit allow/block feedback signal to the user. This binary strategy leaves models vulnerable to adversarial attacks that alter keywords to bypass detection, and it causes high false-alarm rates that degrade the experience for benign users. To address such vulnerabilities, we propose Disciplined Diffusion (DDiffusion), a novel robust text-to-image diffusion that counters Not Safe For Work (NSFW) generation by uncovering implicit malicious semantics in prompt embeddings. DDiffusion leverages a semantic retrieval mechanism to evaluate prompts against concept distributions rather than relying on brittle pairwise similarity. Furthermore, it employs a localization method during the diffusion process to selectively edit only the harmful regions of the generated image. By returning locally sanitized images instead of applying uniform blocking, DDiffusion suppresses malicious content while preserving generation fidelity for benign prompts and avoiding the binary allow-deny signal on which existing probing attacks rely.