Fortinet Acquires Virtue AI to Strengthen Agentic AI Security and Continuous Protection

Fortinet's Virtue AI acquisition rebalances the agentic AI security equation

Fortinet's Virtue AI acquisition rebalances the agentic AI security equation Fortinet Inc. announced its intent to acquire Virtue AI Inc. Monday, and though the financial aspects are interesting, what's more significant is how this will help businesses with their next big security challenge: How do you secure agentic artificial intelligence? Generative AI has raised new concerns about data leakage, unsanctioned use and inaccurate outputs. Agentic AI raises the security stakes because it's significantly harder to secure, and the impact of breaches is greater. These systems can retrieve information, call application programming interfaces, trigger workflows, interact with applications and increasingly act on behalf of people. That makes an AI agent less like a "chatbot" and more like a digital employee with access credentials, permissions and the ability to affect business processes. The industry's response cannot be limited to securing the model or filtering prompts. Organizations need to secure the full lifecycle of autonomous AI: what data an agent can access, which tools it can use, what decisions it is permitted to make, how those actions are monitored and when humans must intervene. That is why Fortinet's acquisition of Virtue AI, which brings AI runtime protection, automated AI validation and agentic AI security capabilities to the company, is strategically important. Why agentic AI changes the security equation Traditional application security was designed for relatively predictable systems. Code may be updated periodically, but it generally operates within defined parameters. Users authenticate, applications make known requests and security teams can establish policies based on expected behavior. Agentic AI operates differently. An agent can dynamically decide which tools to invoke, what information to retrieve, and what next step to take based on the context it receives. It may interact with multiple systems within a single workflow, drawing on customer data, internal knowledge bases, cloud services and business applications. In some cases, one agent may delegate tasks to another. That flexibility is what makes agentic AI valuable. It is also what makes it difficult to secure. The core challenge is that agents operate in continuously changing environments. A model may be updated. A prompt may be modified. A new API connector may be added. The data source for a retrieval system may change. A business team may expand an agent's role from answering questions to initiating transactions. Each change can alter the system's risk profile. This raises several security concerns. Prompt injection can cause an agent to ignore its intended instructions or manipulate it into taking an unintended action. Excessive permissions can turn a minor error into a significant incident. Connected tools can create indirect paths into critical systems. Sensitive data can be exposed through poorly governed retrieval or workflow execution. Because agents can operate at machine speed, a flawed decision can escalate much faster than a traditional human-driven security event. The biggest mistake enterprises can make is treating AI agents as just another application category. They are dynamic actors in the environment. They have identities, permissions, behavioral patterns and connections to other systems. Security controls need to reflect that reality. Why continuous AI protection is needed Most security programs still rely heavily on point-in-time assessments. An application is tested before going into production. A vendor is reviewed during procurement. A configuration is audited periodically. These processes remain important, but they are insufficient for autonomous AI. An AI agent can behave differently tomorrow even if no one changes the underlying business process. A new data set may introduce unexpected content. A model provider may alter the model's behavior. A tool integration may change its capabilities. An adversary may find a new way to influence the agent through malicious content, instructions or inputs. That is why continuous AI protection is so important. Continuous protection means validating an agent before deployment, monitoring it while it operates and reassessing it whenever its model, data, tools, permissions or role changes. It also means moving beyond simple guardrails at the model interface and gaining visibility into the agent's decisions, tool calls and outcomes. The objective should not be to stop enterprises from using agentic AI. That would be unrealistic and strategically counterproductive. The goal should be to enable organizations to deploy agents with confidence by establishing meaningful guardrails for their behavior. A practical, continuous-protection approach should include several elements: * Clear inventory of all AI agents, their owners and their business purpose. * Visibility into the data sources, applications and tools each agent can access. * Least-privilege permissions that limit agents to only the actions they need. * Continuous testing for prompt injection, unsafe tool use, data leakage and policy violations. * Runtime monitoring that detects unusual behavior or actions outside expected parameters. * Human approval requirements for high-impact tasks, such as financial transactions, system changes or access modifications. * Detailed logs that connect prompts, agent decisions, tool calls, identities and outcomes. This is where security will need to evolve from protecting infrastructure and applications to governing behavior. The ability to see that an agent connected to a particular application is useful. The ability to understand why it acted, what data it used, which tool it called and whether the action complied with policy is far more valuable. How Virtue AI fits Fortinet's security platform story Fortinet has long emphasized the convergence of networking and security through its Security Fabric. The company's core argument is that enterprises are better served when security controls share intelligence, policy and enforcement across networks, clouds, endpoints and operations, rather than operating as disconnected point products. The addition of Virtue AI extends that thesis to agentic AI. Fortinet already has AI-related security capabilities, including tools to protect AI interactions and broader efforts to integrate AI into security operations. Virtue AI adds a more specific layer focused on validating and protecting AI agents as they operate. Securing traffic to and from a large language model is necessary, but it is only one part of the problem. Enterprises also need to secure what the agent does after it receives an answer from a model. Can it access sensitive data? Can it send an email? Can it create a ticket? Can it change a configuration? Can it trigger a financial workflow? Can it invoke another agent? Virtue AI's capabilities should help Fortinet broaden its platform story from protecting AI access to protecting AI actions. If integrated effectively, the technology could provide Fortinet customers with a way to apply consistent policy and visibility across the network, applications, endpoints and increasingly autonomous systems operating within those environments. That is especially important for enterprises already struggling with security tool sprawl. Few organizations want to deploy a separate point product for each new AI-related risk. They want AI protection that fits into existing operational workflows, uses existing telemetry and can be managed alongside broader security controls. The proof of this will come from execution. Fortinet will need to demonstrate how Virtue AI integrates with the broader Security Fabric, how customers can operationalize it without adding unnecessary complexity, and how the technology performs across multivendor environments. The most successful AI-security platforms will not be those with the most ambitious marketing claims. They will be those that make autonomous AI safer without slowing the business value it can create. Recommendations for security pros Security leaders should treat agentic AI as a new class of privileged digital worker, not merely another software-as-a-service application or chatbot. Start by inventorying every AI agent already in use, including departmental experiments, and document the data it can access, the tools it can invoke and the actions it is authorized to perform. The foundational controls are the ones most security pros use today. These include least-privilege access, strong identity governance, segmentation, logging and human approval for high-impact actions. The difference is that these controls must be evaluated continuously because an agent's behavior can change as models, prompts, tools and connected data sources evolve. Organizations should also establish a formal AI-agent lifecycle: Validate an agent before production, monitor its behavior at runtime, test it against adversarial inputs, and revoke or reduce permissions when its role changes. Security teams should insist on telemetry that links an agent's prompt, decision, tool call, identity and outcome. Without that end-to-end view, investigating an AI-driven incident becomes an exercise in reconstructing an event after damage has already occurred. Finally, enterprises should favor security architectures that can enforce policy consistently across networks, identities, endpoints, applications and AI interactions, rather than adding another isolated dashboard for each new AI use case. Final thoughts Fortinet's Virtue AI acquisition reflects an important market reality: Agentic AI is likely to create more security work, not less. The enterprises that gain the most from autonomous AI will be those that pair experimentation with disciplined guardrails, continuous validation and clear accountability for agent actions. AI may reshape the security market and force every incumbent to evolve, but it will not eliminate the need for security platforms. If anything, it raises the value of a platform capable of seeing, governing and responding across the increasingly connected environment in which AI agents operate. Zeus Kerravala is a principal analyst at ZK Research, a division of Kerravala Consulting. He wrote this article for SiliconANGLE.

Fortinet Advances Continuous AI Protection with the Acquisition of Virtue AI

Virtue AI will enhance the Fortinet AI-Native Security Fabric with continuous agentic AI validation and runtime protection across the AI lifecycle Fortinet today announced it has acquired Virtue AI, an innovator in AI runtime protection, automated AI validation, and security for autonomous AI systems. The acquisition advances the company's broader Security for AI strategy and vision for securing the agentic enterprise, building on Fortinet's existing AI security solution portfolio, including its FortiGate Hyperscale Firewall. As organizations rapidly deploy AI applications and autonomous agents, their attack surface expands beyond traditional networks, users, endpoints, applications, and cloud workloads. It now includes prompts, models, agents, Model Context Protocol (MCP) tools, application programming interface (API) calls, and AI infrastructure. Organizations need to adopt AI with confidence, keep it secure in production, and trust it behaves as intended. Fortinet met that need earlier this year with FortiAIGate, which safeguards large language models (LLMs) from prompt injections, data leakage, model poisoning, excessive resource consumption, and other emerging AI-specific risks. Virtue AI extends that security to AI models, applications, and agentic systems, from development through runtime, leveraging Virtue AI's Guardian Agent abilities and key product capabilities, including: * Agentic system red-teaming: Tests autonomous agents for exploitable weaknesses across more than 50 sandboxed environments and 14 high-stakes domains, including simulated prompt-injection and MCP-based attacks against leading agent frameworks. * Agent protection, governance, and visibility: Provides full visibility into agents and AI tools running in their environment, discovers unsanctioned AI applications and agents, scans MCP tools and source code for hidden risks, monitors agent behavior, and blocks malicious tool calls before they act. * Continuous AI validation: Identifies new risks across every model update and fine-tuning of policies, while generating audit-ready evidence to support security and compliance reviews. The automated red-teaming runs across hundreds of attack vectors and more than 1,000 risk categories, with multimodal testing and on-demand reporting for security, risk, and compliance teams. * Real-time guardrails: Enforces customizable policies across text, images, video, audio, and AI-generated code to prevent harmful content, sensitive data, jailbreaks, and vulnerable code from reaching users or downstream systems. "AI is fundamentally changing enterprise computing, and security must evolve just as quickly," said Ken Xie, Founder, Chairman of the Board, and Chief Executive Officer at Fortinet. "Virtue AI's technology will advance our vision for continuous AI assurance, helping customers govern and protect AI systems throughout their lifecycle while operating them confidently at enterprise scale." According to Gartner®, "the market for securing AI ecosystems and AI agents is rapidly expanding; products and tools are expected to expand from $2.8 billion in 2026 to $16.4 billion by 2030."[1] Fortinet believes that anticipated market expansion reflects evolving industry demand to secure the AI era. Customers already rely on the Fortinet AI-native Security Fabric for integrated protection across networks, endpoints, clouds, applications, and AI deployments. This acquisition complements FortiAIGate and further strengthens Fortinet's AI runtime security capabilities with Virtue AI's automated validation and real-time protection. Combined with coordinated enforcement and FortiGuard Labs threat intelligence, it will give organizations the confidence to secure AI systems throughout their lifecycle.