Data Collection / Research Resource
14.5k 2026-04-21
jujumilk3/leaked-system-prompts
A curated repository of leaked system prompts from various large language model (LLM) services, valuable for research and understanding LLM behavior.
Core Features
Collects leaked system prompts from popular LLM services.
Supports community contributions with verification requirements.
Emphasizes verifiable sources and reproducible prompts.
Maintains DMCA compliance by avoiding sensitive commercial code.
Cited in academic papers, indicating its research value.
Detailed Introduction
This GitHub repository serves as a critical resource for researchers, developers, and security professionals by compiling leaked system prompts from widely used large language model (LLM) services. It offers insights into the underlying instructions guiding LLM behavior, aiding in the study of prompt engineering, model biases, and potential vulnerabilities. The project encourages community contributions, ensuring data quality through source verification, and strictly adheres to DMCA policies to prevent legal issues, making it a reliable and frequently cited reference in academic research.