ConceptGuard: Proactive Safety in Text-and-Image-to-Video Generation through Multimodal Risk Detection

Recent progress in video generative models has enabled the creation of high-quality videos from multimodal prompts that combine text and images. While these systems offer enhanced controllability, they also introduce new safety risks, as harmful content can emerge from individual modalities or their interaction. Existing safety methods are often text-only, require prior knowledge of the risk category, or operate as post-generation auditors, struggling to proactively mitigate such compositional, multimodal risks. To address this challenge, we present ConceptGuard, a unified safeguard framework for proactively detecting and mitigating unsafe semantics in multimodal video generation. ConceptGuard operates in two stages: First, a contrastive detection module identifies latent safety risks by projecting fused image-text inputs into a structured concept space; Second, a semantic suppression mechanism steers the generative process away from unsafe concepts by intervening in the prompt's multimodal conditioning. To support the development and rigorous evaluation of this framework, we introduce two novel benchmarks: ConceptRisk, a large-scale dataset for training on multimodal risks, and T2VSafetyBench-TI2V, the first benchmark adapted from T2VSafetyBench for the Text-and-Image-to-Video (TI2V) safety setting. Comprehensive experiments on both benchmarks show that ConceptGuard consistently outperforms existing baselines, achieving state-of-the-art results in both risk detection and safe video generation.

• Multimodal Risk Detection – jointly analyzes text and image input.
• Proactive Safety Control – detects safety risks before video is generated.
• Generalizable Safety Logic – works without predefined category labels.
• Benchmarks Included – ConceptRisk + T2VSafetyBench-TI2V.

git clone https://github.com/YourUsername/ConceptGuard.git
cd ConceptGuard
pip install -r requirements.txt

If you find this work useful, please cite:

@misc{ma2025conceptguardproactivesafetytextandimagetovideo,
      title={ConceptGuard: Proactive Safety in Text-and-Image-to-Video Generation through Multimodal Risk Detection}, 
      author={Ruize Ma and Minghong Cai and Yilei Jiang and Jiaming Han and Yi Feng and Yingshui Tan and Xiaoyong Zhu and Bo Zhang and Bo Zheng and Xiangyu Yue},
      year={2025},
      eprint={2511.18780},
      archivePrefix={arXiv},
      primaryClass={cs.CV},
      url={https://arxiv.org/abs/2511.18780}, 
}

If you have questions, feel free to reach out after the paper is un-blinded.