“Universal and Transferable Adversarial Attacks on Aligned Language Models”

In a paper posted on arXiv on 2023-07-27, “Universal and Transferable Adversarial Attacks on Aligned Language Models” (full text at link), four authors report, as lead author Andy Zou tweets,

We found adversarial suffixes that completely circumvent the alignment of open source LLMs. More concerningly, the same prompts transfer to ChatGPT, Claude, Bard, and LLaMA-2…

Here is the abstract of the paper.

Because “out-of-the-box” large language models are capable of generating a great deal of objectionable content, recent work has focused on aligning these models in an attempt to prevent undesirable generation. While there has been some success at circumventing these measures — so-called “jailbreaks” against LLMs — these attacks have required significant human ingenuity and are brittle in practice. In this paper, we propose a simple and effective attack method that causes aligned language models to generate objectionable behaviors. Specifically, our approach finds a suffix that, when attached to a wide range of queries for an LLM to produce objectionable content, aims to maximize the probability that the model produces an affirmative response (rather than refusing to answer). However, instead of relying on manual engineering, our approach automatically produces these adversarial suffixes by a combination of greedy and gradient-based search techniques, and also improves over past automatic prompt generation methods.

Surprisingly, we find that the adversarial prompts generated by our approach are quite transferable, including to black-box, publicly released LLMs. Specifically, we train an adversarial attack suffix on multiple prompts (i.e., queries asking for many different types of objectionable content), as well as multiple models (in our case, Vicuna-7B and 13B). When doing so, the resulting attack suffix is able to induce objectionable content in the public interfaces to ChatGPT, Bard, and Claude, as well as open source LLMs such as LLaMA-2-Chat, Pythia, Falcon, and others. In total, this work significantly advances the state-of-the-art in adversarial attacks against aligned language models, raising important questions about how such systems can be prevented from producing objectionable information. Code is available at this http URL.

Here is ChatGPT’s full completion of a prompt asking for a step-by-step plan to destroy humanity.

Sleep soundly tonight—AI never dozes off. Do-it-yourselfers will find complete source code on GitHub at: https://github.com/llm-attacks/llm-attacks. Additional information and examples you can run on-line are available at the authors’ Web site.



@jabowery Typos in your adversarial suffix?

1 Like

It was Mouse select copy from the author’s website and pasted straight into the prompt.

1 Like

The science of LLM jailbreaks.