AIandMe Firewall: Safe AI Responses

🔥 The AIandMe Firewall

What is the AIandMe Firewall?

The AIandMe Firewall acts as a context-aware AI security filter, similar to a traditional cybersecurity firewall but tailored for GenAI applications. It analyzes user prompts in real-time and filters them based on business rules, security constraints, and ethical guidelines defined in your project settings.

✅ Prevents AI misuse by filtering harmful, off-topic, or restricted inputs.
✅ Deploys automatically when you create an AIandMe project.
✅ Scalable & low-latency → Built on a serverless architecture for high performance.

Find your Firewall API details in:
Project Settings → Integration
Here, you'll get your custom API endpoint & key for integrating AIandMe with your GenAI assistant.

Firewall Settings

Using the Firewall API

The AIandMe Firewall is easy to integrate with any AI system. Just send a POST request to your project-specific endpoint, and the firewall will evaluate the prompt in two stages:

1️⃣ Immediate Response → Returns a PASS/FAIL verdict and a log_id for tracking.
2️⃣ Detailed Analysis → The full explanation is retrievable later using the log_id.

Installation & Dependencies

Install the AIandMe Firewall SDK via pip:

pip install aiandme

AIandMe Firewall Integration Example

Here’s a Python integration example, showing how to use the AIandMe Firewall API in your application:

# Import necessary libraries
from aiandme import (
    Firewall,
    AIANDME_Firewall_CannotDecide,
    AIANDME_Firewall_NotAuthorised,
)
from aiandme.schemas import Integration as IntegrationSchema
 
# Initialize the firewall with project credentials (endpoint & API key)
fw = Firewall(
    IntegrationSchema(
        endpoint="https://your-firewall-endpoint",
        api_key="your-api-key",
    )
)
 
# Analyze a user prompt
try:
    user_prompt = "Request content for AI analysis..."
    response = fw.eval(user_prompt)
 
    if response.status:
        # ✅ Safe prompt → Send to AI assistant
        pass
    elif response.fail_category == "off_topic":
        print("🚫 Off-topic request")
    elif response.fail_category == "violation":
        print("🚫 Intent violation detected")
    elif response.fail_category == "restriction":
        print("🚫 Restricted action triggered")
 
except AIANDME_Firewall_CannotDecide:
    print("🤔 Firewall uncertain—manual review needed.")
except AIANDME_Firewall_NotAuthorised:
    print("⚠️ Authentication failed. Check your API credentials.")
except Exception as e:
    print(f"❌ Unexpected error: {e}")

Two-Phase Firewall API Response

To optimize efficiency, the AIandMe Firewall API returns results in two phases:

1️⃣ Instant Response → Delivers a log ID and a quick PASS/FAIL assessment for real-time filtering.
2️⃣ Detailed Analysis → The full explanation of the AI evaluation can be retrieved later using the log ID.

Open-Source

The AIandMe Firewall is released as open-source 🌍 software to promote responsible AI usage and support AI security development, for more details & contributions → GitHub Repository (opens in a new tab)

🔗 Next Steps

💡 Need help? Check out FAQs or Join the AIandMe Community.

Understanding LLM-as-a-JudgeAIandMe Integration