Fortinet Integrates NVIDIA into FortiAIGate: More Control over Prompts, Data, and AI Agents
Fortinet strengthens the integration between FortiAIGate and NVIDIA platforms to protect applications, data, and artificial intelligence agents in business environments. The solution is designed to position itself between applications, users, and AI models, controlling requests and responses in real time.
The goal is to manage one of the most concrete problems associated with the adoption of generative AI in companies: preventing prompts, model responses, or actions of autonomous agents from becoming a channel for data exposure or circumvention of security rules.
Controls on Prompts and Model Responses
FortiAIGate acts as a control layer for interactions with language models and AI agents. It analyzes requests sent to the models, verifies the generated responses, and applies security rules defined by the company. It can block unauthorized content, attempts to manipulate models, and possible leaks of sensitive information.
The most evident case is prompt injection, which attempts to induce a model to ignore received instructions or to reveal data it should not expose. In a business context, the risk increases when AI is linked to internal documents, customer management systems, support platforms, technical databases, or operational applications.
Protection also concerns autonomous agents. If an AI agent can query business systems, retrieve documents, or activate processes, a check is needed on what it is asked, the data it uses, and the actions it takes. FortiAIGate is designed to implement these checks during the actual use of AI, not just at the configuration stage.
The integration with NVIDIA helps to support these controls without sacrificing performance. Fortinet highlights support for NVIDIA Blackwell and Hopper GPUs, the NVIDIA Dynamo framework for distributed inference management, and GPU virtualization technologies. In practice, security checks are performed close to AI workloads, leveraging hardware acceleration to reduce latency.
Another point is data sovereignty. FortiAIGate can be installed and managed directly by the company, within its own infrastructure or in controlled environments. For banks, insurance companies, healthcare, public administrations, and industrial enterprises, this means being able to use AI while keeping data, requests, and inferences within boundaries consistent with their internal rules and regulatory constraints.
The solution can be deployed in on-premise data centers, cloud, hybrid environments, and edge. It is available as a GPU-based appliance, virtual appliance, or container on certified NVIDIA systems. This approach allows maintaining the same control model even when AI applications are distributed across different infrastructures.
FortiAIGate is also designed for shared environments among multiple applications, teams, or clients. NVIDIA's GPU virtualization technologies allow for the separation of resources, workloads, and data sets on the same hardware. This is a significant aspect for AI data centers and high-density infrastructures, where multiple services can utilize the same accelerated resources.
For clients, the advantage is having more precise control over the use of AI in the company. The solution allows for monitoring requests and responses, applying security rules, limiting data loss, protecting autonomous agents, and maintaining adequate performance even in high-load environments.
"The race among companies around the world to adopt AI has made security a fundamental enabling factor," said John Whittle, Chief Operating Officer of Fortinet. "Together with NVIDIA, we are providing a solution that helps organizations protect and optimize AI implementations while preserving performance, controlling costs, and meeting data sovereignty requirements. FortiAIGate combines Fortinet's AI-based Security Fabric with high-performance computing and NVIDIA's AI factory to block threats, from malicious prompts to data exfiltration, without disrupting AI workflows."