Fortinet Acquired Virtue AI to Protect Enterprises From Their Own AI Agents

Fortinet announced on August 17 that it has acquired Virtue AI, a company built specifically to secure the attack surface that enterprises create when they deploy AI agents.

Fortinet's existing AI security portfolio, led by FortiAIGate launched earlier in 2026, addresses the LLM layer, protecting large language models from prompt injections, data leakage, model poisoning, and excessive resource consumption, according to the company. 

That coverage addresses what happens inside a model. Virtue AI addresses what happens around it, the autonomous agents acting on model outputs, the MCP tools those agents call, the API connections they traverse, and the AI infrastructure they run on.

As organizations deploy agentic AI, their attack surface extends beyond the traditional parameters of enterprise security. Networks, endpoints, cloud workloads, and applications remain relevant. But they are now joined by prompts, agent workflows, MCP tool integrations, and the AI infrastructure layer itself.

"AI is fundamentally changing enterprise computing, and security must evolve just as quickly," said Ken Xie, Founder, Chairman, and CEO of Fortinet.

The four capabilities Virtue AI brings to Fortinet's Security Fabric cover the full lifecycle of an agentic AI deployment, from pre-production testing through runtime protection.

Agentic system red-teaming tests autonomous agents for exploitable weaknesses before they go live, running across more than 50 sandboxed environments and 14 high-stakes domains, including simulated prompt-injection attacks and MCP-based attacks against leading agent frameworks, according to the company. 

The testing surface covers the specific attack patterns that agentic systems face rather than the general cybersecurity attack patterns that traditional red-teaming addresses.

Agent protection, governance, and visibility gives security teams a view into which agents and AI tools are running in their environment, including unsanctioned applications deployed without IT knowledge. 

The system scans MCP tools and source code for hidden risks, monitors agent behavior in real time, and blocks malicious tool calls before they execute, according to the company.

Continuous AI validation addresses the governance gap that emerges every time a model is updated or a policy is fine-tuned, generating audit-ready evidence across hundreds of attack vectors and more than 1,000 risk categories to confirm that security posture has not degraded with each model change. 

Real-time guardrails enforce customizable policies across text, images, video, audio, and AI-generated code to prevent harmful content, sensitive data exposure, jailbreaks, and vulnerable code from reaching users or downstream systems.

Gartner projects the market for securing AI ecosystems and AI agents will grow from $2.8 billion in 2026 to $16.4 billion by 2030.

>

>