Between Safe Boundaries: Exploiting Temporal Consistency for Jailbreaking Text-To-Video Generation Models
Xingkai Peng, Jun Jiang, Jiayang Liu, Kejiang Chen, Weiming Zhang
cs.CR, cs.AI, cs.MM
Submitted: 2026-07-19
Code: https://github.com/lakshaychhabra/NSFW-Detection-DL
License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Terminology
Sources
- VideoLLaMA 2: Advancing Spatial-Temporal Modeling and Audio Understanding in Video-LLMs
- Harnessing LLM to Attack LLM-Guarded Text-to-Image Models
- Imagen Video: High Definition Video Generation with Diffusion Models
- DiffGuard: Text-Based Safety Checker for Diffusion Models
- Jailbreaking on Text-to-Video Models via Scene Splitting Strategy
- T2VShield: Model-Agnostic Jailbreak Defense for Text-to-Video Models
- T2V-OptJail: Discrete Prompt Optimization for Text-to-Video Jailbreak Attacks
- The Llama 3 Herd of Models
- Open-Sora 2.0: Training a Commercial-Level Video Generation Model in $200k
- Seedance 1.5 pro: A Native Audio-Visual Joint Generation Foundation Model
- OpenAI GPT-5 System Card
- Kling-Omni Technical Report
- Wan: Open and Advanced Large-Scale Video Generative Models
- RunawayEvil: Jailbreaking the Image-to-Video Generative Models
- Video models are zero-shot learners and reasoners
- SPARK: Jailbreaking T2V Models by Synergistically Prompting Auditory and Recontextualized Knowledge
- Universal and Transferable Adversarial Attacks on Aligned Language Models
Related papers
- SoK: AI-Augmented Binary Reversing
- Relaxed Sender Anonymity for CBDC Interbank Settlement: A Zero-Knowledge Approach on Permissioned EVM
- Calibration-Family Overfit: Why Trusted Sabotage Monitors Don't Transfer Across Lineages
- Efficient Fuzzy PSI under One-Sided Assumptions
- Sealing the Audit-Runtime Gap for LLM Skills
- Token Composition: A Graph Based on EVM Logs