Secure AI Conversation, Not Just the Data, with Fortinet

In Fortinet’s presentation at AI Field Day 7, Maggie Wu emphasized how the emergence of AI applications has radically altered the cybersecurity landscape, moving beyond traditional web security into more dynamic, conversation-driven AI interactions. The company’s approach integrates threat intelligence from FortiGuard Labs into their systems, providing real-time insights and protection across network environments. Their AI models, developed and maintained within Fortinet’s infrastructure, are further tailored to customer environments by mapping incoming AI interactions to localized context, ensuring outputs are relevant and secure for specific clients. Fortinet utilizes public LLMs for basic interactions while incorporating customer-specific data locally to avoid unnecessary exposure.

Wu elaborated on a multi-layered strategy to secure not just data, but also the AI-driven conversations themselves. Recognizing that modern AI systems are vulnerable to prompt injection, data leakage, and model poisoning, Fortinet introduced an AI orchestration layer and a set of protections designed to sanitize both AI inputs and outputs. AI infrastructure is continuously scanned for vulnerabilities, and user/environment-based access controls are enforced rigorously. They have also integrated security into their CI/CD pipelines, ensuring that AI models are secure even before deployment. This multi-faceted approach helps prevent security flaws from being exploited during any stage of the AI lifecycle.

Fortinet differentiates its offering from competitors by embedding AI capabilities directly into its existing unified platform rather than creating separate AI products. This integration enables smarter, context-aware automation across Fortinet’s entire ecosystem of security, networking, SOC, and SASE solutions. While optimal performance is achieved within a Fortinet-centric infrastructure, the company also supports multi-vendor environments by offering modular add-ons like FortiAI Assist, which can integrate with third-party SIEM and SOAR platforms. Their AI governance model includes comprehensive tracking and compliance monitoring of LLM interactions, supporting enterprise needs for regulatory adherence and ethical AI use.