Frequently Asked Questions

System Prompts Leaks is a comprehensive repository documenting the system prompt instructions for all major AI chatbots. It includes prompts from Anthropic (Claude), OpenAI (ChatGPT), Google (Gemini), xAI (Grok), and many others. The project aims to make these internal instructions transparent and accessible for researchers, developers, and AI enthusiasts.

System prompts are collected through various methods including reverse engineering, API documentation analysis, and community contributions. Each prompt is carefully verified against multiple sources to ensure accuracy. The repository includes evidence of authenticity when available, such as confirmation from AI models themselves or official documentation.

The repository is updated regularly, typically within days of new model releases or significant system prompt changes. Major updates like Claude Fable 5 or GPT-5.5 are documented as soon as they're discovered. The "Recently Updated" section shows the latest additions, with some models receiving multiple updates per month as providers refine their systems.

The repository covers system prompts from over 50 AI models and products, including:

• Anthropic: Claude Fable 5, Opus 4.8, Sonnet 5, Code, Design, Cowork
• OpenAI: GPT-5.5 (Thinking, Instant, API), Codex, various personalities
• Google: Gemini 3.5 Flash, 3.1 Pro, Antigravity CLI
• xAI: Grok 4.3 Beta, Grok Build, Grok Expert
• Microsoft: GitHub Copilot, VS Code Copilot Agent, Copilot for macOS
• And many others including Perplexity, Cursor, Meta AI, Mistral, and more

All content in this repository is licensed under Creative Commons Zero v1.0 Universal (CC0). This means you can freely copy, modify, and distribute the system prompts for any purpose without permission. The license is designed to maximize the sharing and reuse of these valuable AI insights.

Contributions are welcome! You can contribute by:

• Submitting pull requests with new system prompts
• Reporting outdated or incorrect prompts
• Improving documentation and formatting
• Helping verify authenticity of new submissions

Check the repository's contribution guidelines for detailed instructions. All contributions should follow the established format and include verification evidence when possible.

System prompts are the foundational instructions given to AI models before any user interaction. They define the model's behavior, personality, capabilities, and constraints. Unlike regular prompts that users input during conversations, system prompts operate at a deeper level, shaping how the model interprets and responds to all user inputs. Understanding these prompts reveals how AI companies control their models' behavior and capabilities.

The "Recently Updated" section on the main page shows the latest additions. Major updates are also highlighted in the repository's README with direct links to new prompts. For real-time updates, you can:

• Star the repository to receive GitHub notifications
• Follow the project maintainer on X (@asgeirtj)
• Subscribe to repository updates via email
• Check the commit history for daily changes

To stay updated on new model releases:

• Follow official AI company announcements
• Monitor tech news sources like The Washington Post (which has featured this repository)
• Set up GitHub notifications for this repository
• Join the project's community discussions
• Check the "Recently Updated" section regularly for new additions

System prompts are valuable for:

• Researchers studying AI behavior and alignment
• Developers building applications on top of these models
• Ethicists evaluating AI safety measures
• Competitors understanding market positioning
• Educators teaching about AI systems
• Journalists reporting on AI developments
• Users understanding how their AI assistants work