Fortinet launches AI security service for FortiGate users

Bridging the security gap

As organizations integrate large language models (LLMs) and autonomous agents into their daily workflows, traditional security tools are struggling to keep up. Standard security measures typically look for known signatures or malicious IP addresses, but they cannot understand the intent behind a text prompt or a model's response. This creates a vulnerability where sensitive data can leak through interactions with AI tools.

Fortinet is addressing this gap by launching the FortiGuard Inline AI Security Service. This new offering provides specific guardrails designed to inspect AI activity in real time. It aims to stop AI-specific threats that might trigger harmful API calls or autonomous actions from agents. By focusing on the content of the interaction, the service provides a layer of protection that traditional network security lacks.

The goal of this launch is to make advanced security accessible to a wider range of organizations. Instead of requiring companies to build out entirely new infrastructure, Fortinet is delivering these capabilities through a cloud-delivered service. This allows customers to gain deeper visibility into how AI is being used across their business without significant operational complexity.

How the guardrails work

The service functions as a real-time control point between AI applications and the underlying models. It analyzes every interaction based on the specific model, application, user role, and the broader business context. This allows the system to enforce policies dynamically across various types of data, including text, images, audio, video, and code. It is designed to deliver low latency while maintaining a low rate of false positives.

One of the core technical components is the ability to inspect Model Context Protocol (MCP) tools. These are tools that allow models to interact with external data and systems. The service monitors these communications to ensure that an autonomous agent does not perform an unauthorized action. It also identifies 'shadow AI,' which refers to the use of unauthorized AI applications by employees without the knowledge of the IT department.

For data protection, the service utilizes expanded data loss prevention (DLP) policies. It can detect 35 different types of personally identifiable information (PII) and other confidential data. When the system identifies sensitive information, it can alert administrators, deny the request entirely, or redact the information before it ever reaches the AI model. This ensures that private data remains within the organization's control.

Managing costs and compliance

Beyond security, the service provides tools for governance and financial management. Administrators can set specific permissions, budgets, and spend limits for different projects or workspaces. This is particularly important for preventing 'Denial-of-Wallet' attacks, where malicious prompts are designed to consume excessive tokens or GPU compute power. The system will block requests once a spending limit is reached and alert the team as they approach those thresholds.

The service also simplifies the process of staying compliant with various regulations. It includes more than 50 predefined policies that cover data privacy laws and industry-specific security frameworks. Users can apply these policies with a single click to ensure their AI usage meets legal requirements. Furthermore, the system provides detailed logs of every prompt, output, and tool call for auditors to review.

To support different team needs, the service uses dedicated workspaces. These allow individual business units to have their own specific policies, models, and permissions. While each team has its own setup, the organization maintains centralized visibility and control over all activities. This balance allows for flexibility in how different departments use AI while keeping the security team informed.

Deployment and flexibility

Fortinet is offering two primary ways for customers to deploy these capabilities. The FortiGuard Inline AI Security Service is a cloud-delivered option that is easy to enable for existing FortiGate customers. This path is ideal for organizations that want to avoid the headache of managing separate security infrastructure. It allows for a simpler path to adoption by leveraging the hardware the company already owns.

For organizations with stricter requirements regarding data sovereignty or privacy, Fortinet offers a self-hosted option. Customers can deploy FortiAIGate directly in their own environment, whether that is on-premises, in a private cloud, or on Kubernetes with NVIDIA GPUs. This ensures that all prompts, responses, and logs stay entirely within the organization's own infrastructure. This choice allows companies to decide exactly where and how their AI traffic is processed.

The technology powering these features includes recent acquisitions, such as Virtue AI. This integration helps improve the performance and reliability of the guardrails as companies move AI into production environments. By combining these new tools with Fortinet's existing network and identity security, the company aims to provide a unified approach to managing AI risk. This strategy helps organizations scale their AI use quickly while maintaining a consistent security posture.