Agent Smith: A Single Image Can Jailbreak One Million Multimodal LLM Agents Exponentially Fast

A conference paper, Agent Smith, showing that a single image can jailbreak up to one million multimodal LLM agents exponentially fast. The companion cites its simulation of an adversarial image spreading through up to a million multimodal agents. It shows modules spreading through agents in the laboratory, relevant to propagule abundance (P8) and prediction 4.

Cited in

the companion paper (Companion §7)

Used for

P8Prediction 4P8; prediction 4; modules

Strand

Cybersecurity & AI-enabled attack · Prompt injection, agent worms & supply-chain parasites

Cite

@inproceedings{gu2024agent,
  title = {Agent Smith: A Single Image Can Jailbreak One Million Multimodal LLM Agents Exponentially Fast},
  author = {Xiangming Gu and Xiaosen Zheng and Tianyu Pang and Chao Du and Qian Liu and Ye Wang and Jing Jiang and Min Lin},
  year = {2024},
  howpublished = {Proceedings of the 41st International Conference on Machine Learning (ICML)},
  url = {https://arxiv.org/abs/2402.08567},
}