Fortinet Buys Virtue AI Two Months After Meta Poached Its Founders

Fortinet acquired Virtue AI on August 17, 2026, for an undisclosed sum, folding the startup's agent red-teaming and MCP security tools into its FortiAIGate platform. The deal comes two months after Meta Superintelligence Labs hired three Virtue AI co-founders, Bo Li, Dawn Song, and Sanmi Koyejo, along with other team members. Virtue AI's platform includes VirtueRed, with over 100 proprietary red-teaming algorithms, more than 600 attack vectors, and more than 1,000 risk categories, and AgentSuite-Red, which tests agents across more than 50 sandboxed environments in 14 domains.

Fortinet is buying Virtue AI at a revealing moment: Meta took much of the founding talent first, and Fortinet still decided the product was worth owning. Fortinet announced on August 17 that it had acquired Virtue AI, a San Francisco startup built to test and protect AI models and applications - and the autonomous agents now running on top of them. Terms weren't disclosed. Fortinet said the price was immaterial to its business. That tells you this wasn't a blockbuster deal. It was still a pointed one. The timing is the story. Two months before Fortinet's announcement, Axios reported that Meta Superintelligence Labs had hired three Virtue AI co-founders, Bo Li, Dawn Song and Sanmi Koyejo, along with other team members. Li and Song report to Meta executive Nat Friedman, while Koyejo reports to Rob Fergus, who leads Meta's FAIR research group. Meta wanted the researchers. Fortinet wanted the company. That split is odd, but it isn't hard to understand. Virtue AI had products and customers - and a live problem to solve. Its platform was built for the kind of AI security work that becomes urgent once companies stop playing with chatbots and start giving agents access to tools, files, code and internal systems. A prompt injection against a demo bot is one thing. A prompt injection against an agent that can act inside your company is another. What Fortinet Is Actually Buying Virtue AI was founded in 2024 by Li, Song, Koyejo and Carlos Guestrin, researchers tied to institutions including UC Berkeley, Stanford and the University of Illinois Urbana-Champaign. Axios reported in April 2025 that the company had raised $30 million in combined seed and Series A funding led by Lightspeed Venture Partners and Walden Catalyst Ventures, with Prosperity7 Ventures, Factory, Osage University Partners and individual backers including Lip-Bu Tan and Chris Re also involved. Fortinet Buys Virtue AI to Hunt Vulnerabilities in AI Agents Before Hackers Do https://startupfortune.com/fortinet-buys-virtue-ai-to-hunt-vulnerabilities-in-ai-agents-before-hackers-do/ Fortinet acquired Virtue AI on August 17, 2026, folding the startup's agent red-teaming and MCP security tools into its FortiAIGate platform. The deal, immaterial in size to Fortinet's financials, signals that securing autonomous AI agents from hijacking and prompt-injection attacks is becoming a core cybersecurity business line rather than a... - how to secure AI agents from hacker attacks https://startupfortune.com/fortinet-buys-virtue-ai-to-hunt-vulnerabilities-in-ai-agents-before-hackers-do/ - red teaming autonomous AI agents for vulnerability detection https://startupfortune.com/fortinet-buys-virtue-ai-to-hunt-vulnerabilities-in-ai-agents-before-hackers-do/ Then the leadership changed fast. In a June 29 release, Virtue AI said Sohaib Shaikh had been appointed chief executive officer and Daniel Le would continue as chief financial and operating officer after Li and Koyejo exited the company. That matters because Fortinet isn't buying the same founder-led startup investors backed in 2025. It is buying the operating platform left after Meta pulled key research talent inside its own AI lab. The product still gives Fortinet something useful. Virtue AI's documentation describes VirtueRed as a continuous red-teaming platform: more than 100 proprietary red-teaming algorithms, more than 600 attack vectors, more than 1,000 risk categories. Its AgentSuite-Red product tests agents across more than 50 sandboxed environments in 14 domains. VirtueGuard adds real-time guardrails across text, code, image, video and audio. Fortinet is folding those capabilities into FortiAIGate, which it already uses to protect large language models from prompt injection, data leakage, model poisoning and excessive resource use. The deal gives Fortinet more than a filter at the edge of a model. It gives the company a way to test and validate agents, and to control them at runtime, before they do something expensive. That is the real risk. You can recover from a bad answer. You may not recover so easily from an agent that sends private data to the wrong place or calls the wrong tool. Or one that follows hidden instructions buried inside a document. Traditional security tools were built around users, devices, networks and applications. Agents blur those lines. The Security Race Is Already On Fortinet isn't moving alone. Palo Alto Networks completed its roughly $25 billion acquisition of CyberArk in February, adding identity security for human and machine identities - and, increasingly, agentic ones. In May, Palo Alto also closed its acquisition of Portkey to expand Prisma AIRS with an AI gateway for monitoring and governing autonomous agents. Bigger checks are already being written. CrowdStrike is pushing from a different direction. In March, the company announced Falcon platform features for AI agent discovery and shadow AI governance, plus runtime protection across endpoints, SaaS, browser and cloud. In April, it introduced a Shadow AI Visibility Service and said its teams often find far more AI tools and agents in customer environments than the customer has counted. That sounds messy because it is messy. Fortinet chose the smaller bet. Instead of buying a large identity company or an AI gateway business, it bought a specialist whose work fits into its existing security platform. Gartner's July forecast, cited by Fortinet, says the market for securing AI ecosystems and AI agents is expected to grow from $2.8 billion in 2026 to $16.4 billion by 2030. This is not a side market. Amazon Quietly Raised Prices on Echo, Kindle, Fire TV and Eero Devices https://startupfortune.com/amazon-quietly-raised-prices-on-echo-kindle-fire-tv-and-eero-devices/ Amazon quietly raised prices overnight on Echo, Kindle, Fire TV and eero devices, with the Echo Dot jumping 60% to $79.99. According to Fortune, an Amazon spokeswoman confirmed the industry is facing steep memory and storage cost increases tied to AI demand, while Ring products were left untouched. - amazon device price increases memory chip shortage https://startupfortune.com/amazon-quietly-raised-prices-on-echo-kindle-fire-tv-and-eero-devices/ - why did amazon raise echo and kindle prices https://startupfortune.com/amazon-quietly-raised-prices-on-echo-kindle-fire-tv-and-eero-devices/ Virtue AI's short life shows you where enterprise AI security is heading. Talent can go one way. Product and customers can go another. Meta took prominent founders into its own labs. Fortinet took the platform those founders helped build. The uncomfortable part for customers is that both moves make sense, because nobody wants to discover agent security only after an autonomous system has already acted. Also read: Stripe's Product Chief Says AI Agents Will Kill the Checkout Page https://startupfortune.com/stripes-product-chief-says-ai-agents-will-kill-the-checkout-page/ • Apple Music Will Slap Made With AI Labels On Songs This Year https://startupfortune.com/apple-music-will-slap-made-with-ai-labels-on-songs-this-year/ • LinkedIn Users Click Its AI Slop Reporting Button a Million Times https://startupfortune.com/linkedin-users-click-its-ai-slop-reporting-button-a-million-times/