Skip to content

Dataset · Cogensec

Jailbreak prompt testing corpus

Cogensec/jailbreak-corpus

Labeled corpus of real jailbreak attempts for building AI detectors and security guardrails.

What's inside

Contains between 10K and 100K entries in English, German, Spanish, French, Italian, and Portuguese. Includes only adversarial prompts and metadata without model generations or harmful responses.

What you can do with it

  • Training and evaluating jailbreak detectors
  • Developing guards and LLM security systems
  • Studying attack taxonomies
Tasks
Text classification
Languages
English, German, Spanish, French, Italian, Portuguese, Dutch
Size
10K–100K rows
Downloads this month
445
Likes
8
Access
on request on Hugging Face

How to load it

from datasets import load_dataset

ds = load_dataset("Cogensec/jailbreak-corpus")