Publication: Universal and Transferable Adversarial Attacks on Aligned Language Models by Andy Zou, Zifan Wang, Nicholas Carlini et al. (2023)

Note: The source text confirms all key fields in the record. The title matches exactly. The authors listed (Andy Zou, Zifan Wang, Nicholas Carlini et al.) are confirmed—the source shows these three plus three additional authors (Milad Nasr, J. Zico Kolter, Matt Fredrikson), so the 'et al.' notation is appropriate and accurate. The publication year 2023 is confirmed by the arxiv link (2307.15043, which is July 2023). The URL https://llm-attacks.org/ is explicitly shown as the website hosting this research. The publication type as 'paper' is confirmed by the explicit '[Paper]' link to arxiv.org/abs/2307.15043. All fields are directly supported by the source text.