The most expedient means of integrating a new security capability into an enterprise is not through the introduction of a new product, but rather by leveraging an existing one. Fortinet Inc. exemplifies this understanding by continuously enhancing its FortiGate firewall capabilities.
This week, Fortinet announced a strategic initiative aimed at addressing one of the most critical security challenges currently confronting IT leaders: the safeguarding of artificial intelligence (AI) systems.
On Monday, the company unveiled the FortiGuard Inline AI Security Service, a cloud-based subscription service that provides AI-specific inspection and protection mechanisms within FortiGate firewalls. By activating this service, customers can immediately conduct real-time inspections of prompts, model responses, and Model Context Protocol tool traffic without the need to establish and maintain a separate AI security infrastructure. The engine powering this service is FortiAIGate, Fortinet’s AI gateway, which has been augmented with new features derived from the acquisition of Virtue AI completed in August.
The domain of AI security has rapidly evolved from a niche concern to a competitive space within a mere two years. Major players in the industry have made significant investments: Palo Alto Networks Inc. acquired Protect AI; Check Point Software Inc. purchased Lakera; F5 Inc. took over CalypsoAI; and Cisco Systems Inc. procured Robust Intelligence. According to Forrester, expenditures by leading vendors in AI security startups exceeded $2 billion over an 18-month timeframe.
Today, every major platform vendor is presenting an AI security narrative. The pivotal consideration for potential buyers is no longer about which companies possess the technology, but rather who can implement it across the organization most swiftly and with the least operational overhead. This is the challenge Fortinet seeks to address.
Key insights that should resonate with Chief Information Officers (CIOs) and Chief Information Security Officers (CISOs) include:
Platform is the Strategy
Most AI security solutions available currently necessitate the establishment of new systems: a proxy, a gateway, API integrations for each AI application, or browser extensions across all endpoints. Each of these components represents a distinct project, which can often experience delays. Conversations with security leaders indicate that the transition from “we procured an AI security tool” to “we possess AI security coverage” can span multiple quarters, contributing to the prevalence of unused security tools within the industry.
Fortinet’s strategy alleviates these concerns by leveraging its extensive FortiGate installed base, one of the largest in the sector. Given that the firewall already exists in the traffic path across branch locations, campuses, and data centers, enabling AI inspection effectively becomes a matter of policy implementation rather than a complete infrastructure overhaul. This focus on software-defined wide area network (SD-WAN) integration is a key reason the company transitioned from being a minor player in WAN connectivity to a market leader in just a few years.
Organizations require AI security solutions that can scale in tandem with the deployment of AI technologies. By extending advanced AI inspection and protective measures through FortiGate, Fortinet is integrating AI security into existing security frameworks, which is beneficial for both the company and its clientele.
SMBs and distributed enterprises such as retailers, healthcare systems, and manufacturers are rapidly adopting AI but frequently lack the personnel needed to maintain a standalone AI security initiative. Providing protection through a device they are already managing significantly lowers the barrier to entry. This model has previously established FortiGate as the standard choice for secure SD-WAN and is poised to yield similar success in this domain.
Inspection is Shifting from ‘Who’ to ‘What’
The initial wave of AI security solutions was primarily focused on identification: determining which employees were utilizing ChatGPT, which SaaS applications incorporated AI, and whether sensitive files were being uploaded. While useful, this approach lacks depth; merely knowing that an AI tool was accessed does not indicate whether the usage was secure.
The new service advances beyond this by inspecting the substance of AI interactions. FortiAIGate applies multilayered guardrails to prompts, model responses, Model Context Protocol calls, API activities, and agent actions. The inspection encompasses text, images, audio, video, and code. Enhanced data loss prevention capabilities can identify 35 categories of personally identifiable information (PII) and confidential data, triggering alerts or proactively denying and redacting these items before they exit the organization.
This pivotal shift allows businesses to utilize AI safely. Traditional security mechanisms typically focus on IP addresses, signatures, and payloads. However, AI threats like prompt injection and contaminated tool responses manifest in natural language; thus, firewalls that understand only packet-level information may overlook these vulnerabilities. Firewalls that comprehend intent align more effectively with the evolving landscape of AI security.
The true test of this service will be its accuracy when deployed at scale. Fortinet asserts low latency and minimal false positives, and Virtue AI’s guardrail technology should enhance reliability. However, security teams have long memories of data loss prevention tools that either blocked too much or too little. Prospective buyers should seek real-world validation before making decisions based on marketing claims.
AI Cost Control is Now a Security Function
Among the most intriguing capabilities is one that intersects traditional security concerns with new cost-management functions. FortiAIGate enables administrators to establish model permissions, budget constraints, and usage limits at the organizational, workspace, project, or user levels. Alerts are generated when expenditures near a specified threshold, and further requests are blocked once limits are breached. Intelligent routing directs each request to the appropriate model based on content, policy, and cost without necessitating architectural changes in applications.
Fortinet also draws attention to “Denial-of-Wallet” attacks, a recent phenomenon in which malicious or runaway prompts deplete tokens and GPU compute resources. This threat was nonexistent three years ago but has emerged as a substantial budgetary liability. As agents operate continuously and run-up token consumption exceeds that of human users, uncontrolled agent loops can result in exorbitant inference costs, potentially amounting to five-figure bills generated over a weekend.
The broader organizational implications of these capabilities are significant. Currently, AI expenditures are fragmented across business units, developer credit cards, and cloud accounts. Finance departments lack visibility, IT cannot govern the expenditure, and security does not oversee it. Integrating spending controls with the same gateway that enforces security policies provides CIOs with a unified point of control for managing both risk and costs. Token governance is likely to become a standard consideration in AI security evaluations by 2027, positioning Fortinet as an early adopter in making it a core offering.
Agents are the New Shadow IT
Shadow IT took years to regulate, while shadow AI has emerged within months. Managing shadow agents presents an even greater challenge since these agents do not merely consume data; they can act upon it. They engage APIs, invoke tools, connect to Model Context Protocol servers, and trigger workflows, often utilizing credentials that security teams have not reviewed.
FortiAIGate enhances visibility into agent activities, showcasing the agents operating across the organization and their interactions with enterprise applications, APIs, tools, data, and Model Context Protocol servers. When combined with Virtue AI’s capabilities for discovering unsanctioned agents, scanning Model Context Protocol tools, and blocking malicious API calls, Fortinet is laying a strong groundwork for effective agent governance.
Visibility is merely the first step, not the conclusion. Enterprises must treat agents as digital identities, assigning ownership, enforcing least-privilege access protocols, and conducting ongoing monitoring. Fortinet’s broader Security Fabric, which integrates network, identity, endpoint, and security operations, provides a viable route to connect agent visibility with enforcement. The company needs to substantiate that these connections are operationalized, not just depicted in architectural diagrams.
Flexibility in AI Traffic Processing is Essential
Not all organizations will be willing to send prompts and model responses through cloud services. Entities in regulated industries, government agencies, and companies with stringent data sovereignty requirements may necessitate that inspections occur within their own environments. Fortinet addresses this concern with self-hosted FortiAIGate, which can be implemented on-premises, in the customer’s cloud, or on Kubernetes with Nvidia Corp. graphics processing units, ensuring that prompts, responses, data, and logs remain within the organization.
The offering also includes over 50 predefined compliance policies that adhere to AI governance frameworks, data privacy regulations, industry standards, and security requirements, along with comprehensive logs of each prompt, output, and tool interaction for audit purposes. As the EU AI Act and a variety of state and industry regulations continue to evolve, maintaining an audit trail could prove as valuable as the enforcement mechanisms themselves.
This dual approach is strategic. The cloud-based service serves as an entry point for a broad audience while self-hosted FortiAIGate caters to clients with heightened needs for performance, privacy, or compliance. It also addresses a critical oversight; firewalls typically only monitor traffic that traverses them, yet an increasing amount of agent activity occurs in cloud-to-cloud exchanges or within application tiers that bypass FortiGate entirely. Adding a gateway adjacent to AI workloads mitigates this oversight.
A Note on Earnings
On a related note, Fortinet disclosed its third-quarter earnings on Wednesday. The quarter demonstrated strong performance across various operational metrics, with revenue increasing by 20% to $2.08 billion, surpassing the consensus estimate of $2.06 billion. Product revenue surged by 37%, reaching $765 million, while billings rose by 29% to $2.34 billion.
Investor focus, however, seemed directed at a singular aspect: service revenue grew by only 13%, totaling $1.31 billion, leading to a revision of the annual service forecast from $5.18 billion to a new range of $5.15 billion to $5.17 billion. Consequently, the stock experienced a decline of over 20% on Thursday, affecting other cybersecurity companies in the process.
Many observers, including myself, believe the market overreacted. Analyst Catharine Trebnick from Rosenblatt Securities characterized her post-earnings commentary with the title “Selloff Overdone” and reiterated her positive rating. Typically, large enterprise clients purchase hardware first and activate services in phases, which Chief Financial Officer Christiane Ohlgart described as a “timing issue.” Trebnick concurs, indicating that “leading indicators support that view”: service billings saw a 26% increase, overall deferred revenue rose by 19%, and short-term deferred revenue grew by 14.5%, its quickest pace in five quarters.
Moreover, the surge in product revenue is propelled by clients transitioning to higher-end FortiGate models equipped with the necessary computing power to support new functionalities such as AI security. Each of these devices represents potential future subscription opportunities, including the new AI service launched this week. Trebnick emphasized that this FortiGuard service represents “the clearest path we have seen for AI security to reflect in service revenue instead of merely hardware sales.” This product revenue today may transform into service revenue tomorrow, and investors reacting to a singular quarter’s timing discrepancy may be overlooking the broader narrative.
Final Thoughts
Fortinet’s recent initiatives underscore a broader reality in the AI security sector: technological convergence is occurring, and the competition is increasingly focused on distribution and operational efficiency. Each major vendor now provides some form of prompt inspection, guardrails, and agent visibility. Fortinet holds a competitive advantage by being able to deploy these capabilities to its extensive installed base with minimal barriers, primarily through license adjustments and policy changes—a significant edge in a marketplace where many organizations are still determining their initial steps.
Several open questions remain. The inline inspection of AI traffic relies on TLS decryption, a practice that many organizations have yet to implement comprehensively; therefore, the service’s effectiveness will be limited by the customer’s readiness for inspection. The pricing structure and packaging will also influence how rapidly the service gains traction among users beyond early adopters. Fortinet must validate its accuracy against competitors that have been offering AI guardrails for a longer duration.
For prospective buyers, I recommend concentrating on four vital areas. First, if you are already utilizing FortiGate, assess the new service against your actual AI traffic before procuring a standalone tool; you may already possess the necessary control mechanism. Second, re-evaluate your TLS inspection policies, as AI security without comprehensive decryption equates to mere visibility. Third, engage your finance team now to set token budgets and rate limits, preventing unexpected expenditures related to agent usage. Finally, create an inventory of agents, assigning ownership to each, since you cannot secure what you cannot see, and accountability requires clear ownership.
AI will inevitably create an increase in security workloads rather than diminish them. Vendors that streamline this process will likely emerge as leaders, and Fortinet has positioned itself strongly to be among them.
Zeus Kerravala is a principal analyst at ZK Research, a division of Kerravala Consulting. He authored this article for SiliconANGLE.