Demoing FortiAIGate

Demoing the FortiAIGate

In this section we will show how the FortiAIGate enforces, monitors, and protects LLM communications for customers. We will show the process using a three step process:

• Expose - demonstrate the vulnerability on an unprotected LLM with real attack payloads
• Configure - apply the FortiAIGate control (AI Guard scanner + action)
• Validate - confirm the action with the same payload, view logs, and occasionally run a negative test to verify existing operations

Each phase will produce log evidence (subtype=ai-security, action=blocked) suitable for compliance audits and security posture reporting.

How it Works

FortiAIGate uses AI Guards to look at the data as it flows from the client to the LLM as well as the response from the LLM to the client. These include the following:

Input Guards

This is the information that flows from the client application, in our case the chatbot application. These include:

• Prompt Injection Detection
• Data Loss Prevention
• Toxicity Filtering
• Custom Rule Filters

Output Guards

This is the information that flows from the LLM to the client application. These include:

• Data Loss Prevention
• Toxicity Filtering
• Custom Rule Filters

Let’s Get Started with our Use Cases

Let’s get started!