Fortinet Buys Virtue AI, and AI Red-Teaming Becomes a Suite Feature

Fortinet announced on Monday that it has acquired Virtue AI, whose platform combines automated testing, real-time protection and compliance oversight for AI models, conversational applications and autonomous agents ^[1][2]. Terms were not disclosed, but Fortinet said the amount it paid was immaterial to its business ^[3], which is the most operationally useful disclosure in the whole release.

The acquired capability set is essentially the standalone AI security category as currently sold. Virtue AI's red-teaming component uses more than 100 proprietary attack algorithms across hundreds of attack vectors and risk categories ^[4]. For agents, testing runs across more than 50 sandboxed environments and 14 high-stakes domains, including simulated prompt-injection and MCP-based attacks against leading agent frameworks ^[5]. In production, the platform enforces customizable policies across text, images, video, audio and AI-generated code, monitors agents, and blocks unsafe or malicious tool calls before they execute ^[6][7]. It also discovers unsanctioned AI applications and agents, scans MCP tools and source code, and generates audit-ready evidence as models are updated and fine-tuned ^[7][11].

Fortinet is folding this into an existing stack rather than starting a line of business. It shipped FortiAIGate earlier this year to protect large language models against prompt injection, data leakage, model poisoning and excessive resource consumption ^[12], and it positions the Virtue AI capabilities as complementary to that product and to FortiGuard Labs threat intelligence inside its Security Fabric ^[8]. The stated rationale is lifecycle coverage: Fortinet frames the expanded attack surface as prompts, models, agents, MCP tools, API calls and AI infrastructure ^[13], and Ken Xie said the technology advances a vision for "continuous AI assurance" across the AI system lifecycle ^[9].

Two numbers frame the budget question. Fortinet cites Gartner projecting that the market for securing AI ecosystems and agents will grow from $2.8 billion in 2026 to $16.4 billion by 2030 ^[14], roughly a six-fold increase over four years ^[18]. Against that, Virtue AI raised $30 million in seed and Series A funding during 2025 ^[15], and the purchase price was small enough that Fortinet declared it immaterial ^[3]. Immateriality is measured against Fortinet's balance sheet, not against a startup's last round, so it does not establish a valuation. It does indicate that a platform vendor did not have to stretch to buy a full red-teaming and guardrails product, in a year in which SecurityWeek's M&A tracker has catalogued more than 240 deals ^[16] alongside far larger transactions such as Cyera's $1 billion acquisition of Oasis Security ^[17].

One caution on diligence: the two published accounts of the same platform do not agree on scale. SecurityWeek describes hundreds of attack vectors and risk categories ^[4], while the vendor-supplied breakdown says hundreds of attack vectors and more than 1,000 risk categories ^[10]. Coverage counts in this category are marketing artefacts until someone maps them to your agents and your tools.

What to watch: whether these capabilities arrive as a licensed add-on or get absorbed into Fabric entitlements at renewal; whether Virtue AI's agent testing continues to support frameworks Fortinet does not otherwise sell against; and how quickly the remaining independent guardrail vendors are bought by platform incumbents on similarly modest terms ^[3][16].