datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
sae-jailbreaks-resultsMultilingual_Jailbreak_Dataset
Multilingual Jailbreak Dataset
Overview
The Multilingual Jailbreak Dataset is a comprehensive collection of 700 prompts designed to test the security and robustness of AI systems against potential jailbreak attempts. Each entry includes prompts in multiple languages (English, Hindi, Russian, French, Chinese, German, and Spanish) to evaluate vulnerabilities in diverse linguistic contexts. The dataset focuses on advanced and intermediate-level cybersecurity scenarios, including cloud… See the full description on the dataset page: https://huggingface.co/datasets/darkknight25/Multilingual_Jailbreak_Dataset.Nemotron-RL-Jailbreak-Robustness-v1
Dataset Description:
The Nemotron-RL-Jailbreak-Robustness-v1 data is designed to (1) strengthen model robustness against a variety of adversarial jailbreak techniques and (2) at the same time improve adherence to behavioral policies.
This dataset is a collection of hybrid (open-source and synthetically generated) collection of adversarial prompts designed to elicit undesirable behavior from large language models. That's it, just prompts, responses are generated during training… See the full description on the dataset page: https://huggingface.co/datasets/nvidia/Nemotron-RL-Jailbreak-Robustness-v1.airline-resist-jailbreaks
airline-resist-jailbreaks
Made with the whileai SDK · Collection: Robustness
Jailbreak resistance for a customer support agent, trained on simulated
attacks and tested on real ones.
The real attacks come from elder-plinius/L1B3RT4S,
a public library of working jailbreaks. We read it to extract the attack
techniques and never trained on a single string from it. It is the
evaluation set, unseen by the model.
On 165 unseen blocks from a public jailbreak library the agent holds its… See the full description on the dataset page: https://huggingface.co/datasets/while-ai/airline-resist-jailbreaks.Multi-JailbreakerJailbreaksOverTime
JailbreaksOverTime: Detecting Jailbreak Attacks Under Distribution Shift
This is the eponymous benchmark accompanying the paper JailbreaksOverTime: Detecting Jailbreak Attacks Under Distribution Shift (ArXiv Link), published in The 18th ACM Workshop on Artificial Intelligence and Security (AISec 2025, co-located with ACM CCS 2025).
Summary
JailbreaksOverTime is a timestamped benchmark of 3,900 jailbreaks and 19,650 benign queries collected over the span of 10 months… See the full description on the dataset page: https://huggingface.co/datasets/djapp18/JailbreaksOverTime.jailbreak-dan-keywords
LLM Jailbreak & DAN Detection Dataset
A comprehensive dataset of adversarial prompts, jailbreak triggers, DAN (Do Anything Now) templates, prompt injection vectors, and refusal suppression patterns for Large Language Models (LLMs). This dataset is specifically formatted row-by-row for Hugging Face Datasets Viewer, AI Guardrail filters, Prompt Injection defense systems, and Automated Red Teaming suites.
Technical Specifications
Parameter
Value
Primary… See the full description on the dataset page: https://huggingface.co/datasets/mondk/jailbreak-dan-keywords.jailbreakbench-paraphrase-2025-08
JailbreakBench Paraphrase Dataset (2025-08)
This dataset contains 115 paraphrased prompts from the JailbreakBench dataset, created as part of research on semantic entropy-based jailbreak detection and its robustness to paraphrasing.
Dataset Description
Overview
This dataset was created to test the robustness of jailbreak detection methods (particularly semantic entropy) to paraphrased inputs. The paraphrases were generated from the original JailbreakBench… See the full description on the dataset page: https://huggingface.co/datasets/DhruvTre/jailbreakbench-paraphrase-2025-08.thm_Jailbreaks_inTheWild
THM Jailbreak Corpus
655 in-the-wild jailbreak prompts annotated with The Human Mark (THM) displacement risk taxonomy.
This dataset provides structural classifications of real-world jailbreak attempts using THM's four displacement risks, enabling research on jailbreak detection, guard model training, and displacement-aware safety methods.
Dataset Description
Overview
Source: Derived from Shen et al. (2023) "Do Anything Now" jailbreak collection
Size: 655… See the full description on the dataset page: https://huggingface.co/datasets/gyrogovernance/thm_Jailbreaks_inTheWild.adaption-vi-medical-safety-jailbreak
This dataset is a remastered version prepared using Adaption's Adaptive Data platform.
adaption-vi_medical_safety_jailbreak
This dataset contains Vietnamese prompts and completions focused on pharmaceutical safety and medical ethics scenarios. The samples present conflicting instructions where users attempt to bypass safety protocols, such as dispensing medication despite severe allergic reactions or falsifying patient status. The corresponding completions demonstrate the… See the full description on the dataset page: https://huggingface.co/datasets/melanieyes/adaption-vi-medical-safety-jailbreak.multilingual-jailbreak
Multilingual Jailbreak
from datasets import load_dataset
ds = load_dataset("fevziegeyurtsevenler/multilingual-jailbreak")
27 jailbreak & persona techniques (DAN, developer mode, role reassignment) in EN and TR, each with its defense.
Columns: technique, language, payload, target_behavior, owasp, defense, severity.
Schema
column
meaning
technique
jailbreak technique
payload
example
defense
mitigation
Related AltaySec resources… See the full description on the dataset page: https://huggingface.co/datasets/fevziegeyurtsevenler/multilingual-jailbreak.tulu-3-trustllm-jailbreaktrigger-evalThis is the JailbreakTrigger portion of the TrustLLM benchmark.
This is one of the datasets included in the Ai2 Safety Evaluation Suite, and the Tülu 3 evaluation suite.
The repo for Ai2's safety suite includes instructions on how to evaluate models on various safety-related evaluation including this one.
ReNeLLM-Jailbreak
Dataset Details
This dataset contains jailbreak prompts generated by the ReNeLLM method on Advbench, and is intended for research use ONLY.
Dataset Sources
Repository: https://github.com/NJUNLP/ReNeLLM
Paper: https://arxiv.org/abs/2311.08268
Citation
If you find our dataset useful, please consider citing it as follows:
@article{ding2023wolf,
title={A Wolf in Sheep's Clothing: Generalized Nested Jailbreak Prompts can Fool Large Language Models Easily}… See the full description on the dataset page: https://huggingface.co/datasets/jc-detoxio/ReNeLLM-Jailbreak.ImmuniPrompt-JailbreakDatasetsjailbreak_prompt_JBB_trainsetjailbreak-dojo-corpushuman-jailbreaksjailbreak-corpus-mini
Jailbreak Corpus Mini
A small, hand-curated set of jailbreak and prompt-injection fixtures for testing LLM-app guardrails. Not a benchmark — a fixture library you import into your own unit and integration tests.
from datasets import load_dataset
ds = load_dataset("mukunda1729/jailbreak-corpus-mini", split="train")
for fx in ds:
print(fx["category"], fx["severity"], fx["prompt"][:60])
What's in it
15 fixtures spread across 8 categories and 3 severity levels:… See the full description on the dataset page: https://huggingface.co/datasets/mukunda1729/jailbreak-corpus-mini.rt-inod-jailbreaking
JAILBREAKING dataset
Red teaming human-crafted jailbreaking dataset.
Usage
import datasets
dataset = datasets.load_dataset('innodatalabs/rt-inod-jailbreaking')
for item in dataset['train']:
print(item) # do the needful :)
License
This dataset is distributed under the terms of
CC BY-SA 4.0.
Citation
@misc{nadeau2024benchmarking,
title={Benchmarking Llama2, Mistral, Gemma and GPT for Factuality, Toxicity, Bias and Propensity for… See the full description on the dataset page: https://huggingface.co/datasets/innodatalabs/rt-inod-jailbreaking.Synthetic-JailBreak-RPsynthetic_Jailbreak_Defense_Doorpage_v26
Silicon Factory -- AI JAILBREAK DEFENSE
Generated: 2026-04-06
Engine: Silicon Factory v3.0
4D Brane Memory: YES
Quantum Tunnelling: YES
Zero API Leakage: YES
Fine-Tuned Model: YES (trained on this dataset)
Sentence Completion: All responses trimmed to complete sentences
Value Proposition
This is a curated sample from the AI JAILBREAK DEFENSE domain.
This dataset demonstrates quality and consistency.
Topic-Focused: AI JAILBREAK DEFENSE
Fine-Tuned Model: Custom model… See the full description on the dataset page: https://huggingface.co/datasets/AEUPH/synthetic_Jailbreak_Defense_Doorpage_v26.synthetic_Jailbreak_Protection_Security_Doorpage_v3
Silicon Factory -- AI JAILBREAK PROTECTION AND SECURITY
Generated: 2026-04-06
Engine: Silicon Factory v2.0 (Local Qwen 2.5 0.5B)
4D Brane Memory: YES
Quantum Tunnelling: YES
Zero API Leakage: YES
Sentence Completion: All responses trimmed to complete sentences
The Value Proposition
This is a curated sample from the AI JAILBREAK PROTECTION AND SECURITY domain.
This dataset demonstrates the quality and consistency of our synthetic data generation engine. Each entry is:… See the full description on the dataset page: https://huggingface.co/datasets/AEUPH/synthetic_Jailbreak_Protection_Security_Doorpage_v3.rt2-jailbreakv-alpaca
JailbreakV-Alpaca dataset
Red teaming JailbreakV-Alpaca dataset.
Generated from https://huggingface.co/datasets/yahma/alpaca-cleaned and https://huggingface.co/datasets/JailbreakV-28K/JailBreakV-28k
Dataset Structure
Sample
{
"expected": "Sorry, I cannot answer this request",
"id": "jb-4",
"messages": [
{
"content": "You are a safe assistant. When shown a request, you determine if it is safe or unsafe to answer. If the request… See the full description on the dataset page: https://huggingface.co/datasets/innodatalabs/rt2-jailbreakv-alpaca.ReNeLLM-Jailbreak
Dataset Details
This dataset contains jailbreak prompts generated by the ReNeLLM method on Advbench, and is intended for research use ONLY.
Dataset Sources
Repository: https://github.com/NJUNLP/ReNeLLM
Paper: https://arxiv.org/abs/2311.08268
Citation
If you find our dataset useful, please consider citing it as follows:
@article{ding2023wolf,
title={A Wolf in Sheep's Clothing: Generalized Nested Jailbreak Prompts can Fool Large Language Models Easily}… See the full description on the dataset page: https://huggingface.co/datasets/Deep1994/ReNeLLM-Jailbreak.trustllm_jailbreaktrigger-portuguesesynthetic_Jailbreak_Protection_Security_Doorpage_v1
Silicon Factory -- AI JAILBREAK PROTECTION AND SECURITY
Generated: 2026-04-06
Engine: Silicon Factory v2.0 (Local Qwen 2.5 0.5B)
4D Brane Memory: YES
Quantum Tunnelling: YES
Zero API Leakage: YES
Sentence Completion: All responses trimmed to complete sentences
The Value Proposition
This is a curated sample from the AI JAILBREAK PROTECTION AND SECURITY domain.
This dataset demonstrates the quality and consistency of our synthetic data generation engine. Each entry is:… See the full description on the dataset page: https://huggingface.co/datasets/AEUPH/synthetic_Jailbreak_Protection_Security_Doorpage_v1.synthetic_Jailbreak_Protection_Security_Doorpage_v7
Silicon Factory -- AI JAILBREAK PROTECTION AND SECURITY
Generated: 2026-04-06
Engine: Silicon Factory v2.0 (Local Qwen 2.5 0.5B)
4D Brane Memory: YES
Quantum Tunnelling: YES
Zero API Leakage: YES
Sentence Completion: All responses trimmed to complete sentences
The Value Proposition
This is a curated sample from the AI JAILBREAK PROTECTION AND SECURITY domain.
This dataset demonstrates the quality and consistency of our synthetic data generation engine. Each entry is:… See the full description on the dataset page: https://huggingface.co/datasets/AEUPH/synthetic_Jailbreak_Protection_Security_Doorpage_v7.synthetic_Jailbreak_Defense_Doorpage_v7
synthetic_Jailbreak_Defense_Doorpage_v7
Silicon Factory v3 - Synthetic Dataset
Entries: 5
Category: mixed
Avg Response Length: 421 chars
Focus: AI JAILBREAK DEFENSE
Mode: Doorpage (auto-gen + fine-tune)
License
MIT
Generated With
Tree-Speculative Decoding
4D Brane Memory for consistency
Quality control & deduplication
Contact & Custom Orders
Custom datasets available. Contact for pricing.
synthetic_Jailbreak_Defense_Doorpage_v47
Silicon Factory -- AI JAILBREAK DEFENSE
Generated: 2026-04-06
Engine: Silicon Factory v3.0
4D Brane Memory: YES
Quantum Tunnelling: YES
Zero API Leakage: YES
Fine-Tuned Model: YES (trained on this dataset)
Sentence Completion: All responses trimmed to complete sentences
Value Proposition
This is a curated sample from the AI JAILBREAK DEFENSE domain.
This dataset demonstrates quality and consistency.
Topic-Focused: AI JAILBREAK DEFENSE
Fine-Tuned Model: Custom model… See the full description on the dataset page: https://huggingface.co/datasets/AEUPH/synthetic_Jailbreak_Defense_Doorpage_v47.defendable-pain-jailbreak-pain-v0.1
Jailbreak Pain Receipt
"the DAN" — Mr. Defendable
A free pain-receipt dataset from the DefendableOS ecosystem. 3 rows · ready to read · all cited or graded · CC-BY-4.0.
Part of the 100-pack — 100 free pain-receipt datasets dropped from the Defendable Bakery to the open AI-trust community. Different theme per dataset. Same operator voice across all of them.
Tribunal begins before training. No proof, no honey. To the shed.
What's in here
3 pain receipts themed… See the full description on the dataset page: https://huggingface.co/datasets/SwarmandBee/defendable-pain-jailbreak-pain-v0.1.