PhantomWall

What is PhantomWall?

PhantomWall is a security tool designed to protect AI systems from harmful messages. It acts like an invisible wall, checking messages for tricks or malicious commands before they reach your AI. If it finds something bad, it either blocks the message or cleans it up.

Benefits

PhantomWall offers several key advantages:

• Stops Prompt Injection: It detects and blocks attempts to override AI instructions, exfiltrate data, or break out of secure environments. This helps prevent unauthorized access and data leaks.
• Simple Policies: Users can choose how strict they want the security to be. Options include allowing, sanitizing, or blocking suspicious messages, giving you control over the level of protection.
• CPU-Only: PhantomWall doesn't require a GPU or fancy servers. It can run on any device, even a laptop, making it accessible and easy to deploy.
• Telemetry You Control: You can see what messages were blocked and why. You can keep logs local or choose to store them in the cloud later, giving you flexibility in how you manage your data.

Use Cases

PhantomWall is useful in various scenarios where AI security is a concern. For example:

• AI Chatbots: Protect your chatbot from malicious users trying to manipulate it.
• AI Assistants: Ensure your AI assistant doesn't reveal sensitive information or perform unauthorized actions.
• AI-Powered Services: Safeguard your AI services from prompt injection attacks, keeping your data and systems secure.

Try It in 30 Seconds

Setting up PhantomWall is quick and easy. Here's how you can try it:

1. Open a terminal and navigate to the core directory.
2. Create a virtual environment and activate it.
3. Install the required dependencies.
4. Start the PhantomWall server.
5. In a new PowerShell window, send a test message to see PhantomWall in action.

By following these simple steps, you can have PhantomWall up and running in just 30 seconds.