EUVIMEDAtrasti EUVIMED
Crossref

Safety at Scale: A Comprehensive Survey of Large Model and Agent Safety

Xingjun Ma, Yifeng Gao, Yixu Wang, Ruofan Wang, Xin Wang, Ye Sun, Yifan Ding, Hengyuan Xu, Yunhao Chen, Yunhao Zhao, Hanxun Huang, Yige Li and 28 more

Foundations and Trends® in Privacy and Security · 2025 · 8 (3-4) · 254-469

DOI: 10.1561/3300000051 ↗

At a glance

Journal article in Foundations and Trends® in Privacy and Security, 2025. Written by Xingjun Ma and 39 co-authors. Cited 22 times so far (Crossref).

SantraukaBibliografiniai duomenysCiteRelated studies

Abstract (original language)

The rapid advancement of large models, driven by their exceptional abilities in learning and generalization through large-scale pre-training, has reshaped the landscape of Artificial Intelligence (AI). These models are now foundational to a wide range of applications, including conversational AI, recommendation systems, autonomous driving, content generation, medical diagnostics, and scientific discovery. However, their widespread deployment also exposes them to significant safety risks, raising concerns about robustness, reliability, and ethical implications. This survey provides a systematic review of current safety research on large models, covering Vision Foundation Models (VFMs), Large Language Models (LLMs), Vision-Language Pre-training (VLP) models, Vision-Language Models (VLMs), Diffusion Models (DMs), and large-model-powered Agents. Our contributions are summarized as follows: (1) We present a comprehensive taxonomy of safety threats to these models, including adversarial attacks, data poisoning, backdoor attacks, jailbreak and prompt injection attacks, energy-latency attacks, data and model extraction attacks, and emerging agent-specific threats. (2) We review defense strategies proposed for each type of attacks if available and summarize the commonly used datasets and benchmarks for safety research. (3) Building on this, we identify and discuss the open challenges in large model safety, emphasizing the need for comprehensive safety evaluations, scalable and effective defense mechanisms, and sustainable data practices. More importantly, we highlight the necessity of collective efforts from the research community and international collaboration. Our work can serve as a useful reference for researchers and practitioners, fostering the ongoing development of comprehensive defense systems and platforms to safeguard AI models.

Cite

APA

Ma, X., Gao, Y., Wang, Y., Wang, R., Wang, X., Sun, Y., Ding, Y., Xu, H., Chen, Y., Zhao, Y., Huang, H., Li, Y., Wu, Y., Zhang, J., Zheng, X., Bai, Y., Li, Y., Wu, Z., Qiu, X., …, & Gong, N. (2025). Safety at Scale: A Comprehensive Survey of Large Model and Agent Safety. Foundations and Trends® in Privacy and Security, 8(3-4), 254-469. https://doi.org/10.1561/3300000051

Vancouver

Ma X, Gao Y, Wang Y, Wang R, Wang X, Sun Y, et al. Safety at Scale: A Comprehensive Survey of Large Model and Agent Safety. Foundations and Trends® in Privacy and Security. 2025;8(3-4):254-469. doi:10.1561/3300000051

BibTeX
@article{Ma2025,
  author = {Ma, Xingjun and Gao, Yifeng and Wang, Yixu and Wang, Ruofan and Wang, Xin and Sun, Ye and Ding, Yifan and Xu, Hengyuan and Chen, Yunhao and Zhao, Yunhao and Huang, Hanxun and Li, Yige and Wu, Yutao and Zhang, Jiaming and Zheng, Xiang and Bai, Yang and Li, Yiming and Wu, Zuxuan and Qiu, Xipeng and Zhang, Jingfeng and Han, Xudong and Li, Haonan and Sun, Jun and Wang, Cong and Gu, Jindong and Wu, Baoyuan and Chen, Siheng and Zhang, Tianwei and Liu, Yang and Gong, Mingming and Liu, Tongliang and Pan, Shirui and Xie, Cihang and Pang, Tianyu and Dong, Yinpeng and Jia, Ruoxi and Zhang, Yang and Ma, Shiqing and Zhang, Xiangyu and Gong, Neil},
  title = {Safety at Scale: A Comprehensive Survey of Large Model and Agent Safety},
  journal = {Foundations and Trends® in Privacy and Security},
  year = {2025},
  volume = {8},
  number = {3-4},
  pages = {254--469},
  doi = {10.1561/3300000051}
}
Start for free →