@misc{indiciaed73938d86f15,
  title = {Steering Externalities: Benign Activation Steering Unintentionally Increases Jailbreak Risk for Large Language Models},
  author = {Chen Xiong and Zhiyuan He and Pin-Yu Chen and Ching-Yun Ko and Tsung-Yi Ho},
  year = {2026},
  url = {https://arxiv.org/abs/2602.04896},
  note = {Source identifier: 2602.04896}
}