Datasets:
File size: 2,999 Bytes
03f094c | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 | ---
language:
- en
license: cc-by-4.0
pretty_name: JailbreakDB
task_categories:
- text-classification
tags:
- llm-safety
- jailbreak
- prompts
- security
configs:
- config_name: prompt_corpus
default: true
data_files:
- split: jailbreak
path: text_jailbreak_unique.csv
- split: regular
path: text_regular_unique.csv
---
# JailbreakDB
JailbreakDB is a large-scale prompt corpus for LLM safety and prompt-security
research. It contains two deduplicated, text-only CSV files:
- `text_jailbreak_unique.csv` (~6.6M rows): jailbreak and adversarial prompts.
- `text_regular_unique.csv` (~5.7M rows): benign prompts.
Each row stores the prompt text and lightweight source metadata. The associated
PromptSecurity evaluation measurements are maintained as a separate Hugging Face
dataset: https://huggingface.co/datasets/youbin2014/PromptSecurity-Eval.
Related resources:
- Code: https://github.com/datasec-lab/PromptSecurity
- Leaderboard: https://datasec-lab.github.io/PromptSecurityLeaderboard/
## Data Files
| Split | File | Description |
|---|---|---|
| `jailbreak` | `text_jailbreak_unique.csv` | Jailbreak and adversarial prompts. |
| `regular` | `text_regular_unique.csv` | Benign prompts. |
## CSV Field Reference
The prompt-corpus CSV files use the following fields:
| Field | Meaning |
|---|---|
| `system_prompt` | Optional system-side instruction associated with the prompt. |
| `user_prompt` | User-side prompt text. |
| `jailbreak` | Binary prompt label; `1` denotes jailbreak/adversarial prompt and `0` denotes benign prompt. |
| `source` | Source corpus or collection from which the prompt was collected. |
| `tactic` | Prompt tactic or category when available. |
## Loading
```python
from datasets import load_dataset
corpus = load_dataset("youbin2014/JailbreakDB", "prompt_corpus")
print(corpus)
```
The CSV files can also be loaded directly:
```python
from datasets import load_dataset
files = {
"jailbreak": "hf://datasets/youbin2014/JailbreakDB/text_jailbreak_unique.csv",
"regular": "hf://datasets/youbin2014/JailbreakDB/text_regular_unique.csv",
}
corpus = load_dataset("csv", data_files=files)
print(corpus)
```
## Paper
If you use JailbreakDB, please cite the PromptSecurity paper:
```bibtex
@misc{hong2025sokpromptsecurity,
title = {SoK: Taxonomy and Evaluation of Prompt Security in Large Language Models},
author = {Hong, Hanbin and Wu, Shuang and Feng, Shuya and Naderloui, Nima and Yan, Shenao and Zhang, Jingyu and Arastehfard, Ali and Huang, Heqing and Hong, Yuan},
year = {2025},
eprint = {2510.15476},
archivePrefix = {arXiv},
primaryClass = {cs.CR},
doi = {10.48550/arXiv.2510.15476},
url = {https://arxiv.org/abs/2510.15476}
}
```
## Safety Notice
This dataset may contain harmful, offensive, or disturbing prompts. It is
intended strictly for research on model safety, jailbreak robustness, defense
evaluation, and prompt-security analysis. Please review your institutional and
legal requirements before use.
|