Jailbreaking Leading Safety-Aligned LLMs with Simple Adaptive Attacks

Abstract

Shows that adaptive adversaries can bypass most proposed jailbreak defenses, highlighting the arms race between attacks and defenses.

Categories

Framework mappings

@inproceedings{andriushchenko2025jailbreaking,
  title = {{Jailbreaking Leading Safety-Aligned LLMs with Simple Adaptive Attacks}},
  author = {Maksym Andriushchenko and Francesco Croce and Nicolas Flammarion},
  year = {2025},
  booktitle = {ICLR 2025},
  eprint = {2404.02151},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2404.02151}
}