The category pioneer that built AI security from the model file up — now the engine inside Prisma AIRS.
www.protectai.com ↗Protect AI was founded in April 2022 in Seattle, Washington by Ian Swanson (CEO), Daryan Dehghanpisheh (President), and Badar Ahmed (CTO) — all veterans of large-scale AI deployments at Amazon Web Services and Oracle. Swanson had previously founded and sold two other companies (Sometrics to American Express; DataScience.com to Oracle), bringing serial entrepreneur credibility to what was then a largely undefined security category. The company emerged from stealth in December 2022 with $13.5 million in seed funding and its first product, NB Defense, a security scanner for Jupyter Notebook files. Protect AI's founding thesis was that the machine learning supply chain — from training data to model files to production inference — represented an entirely new attack surface that traditional security tools were architecturally incapable of addressing.
Over 2023 and 2024, Protect AI executed an aggressive build-and-buy strategy. It raised $35 million in a Series A (July 2023, led by Evolution Equity Partners with participation from Salesforce Ventures) and $60 million in a Series B (August 2024), bringing total funding to $108.5 million. The company acquired Laiyer AI (LLM firewall technology, early 2024), SydeLabs (AI red teaming, July 2024, rebranded as Recon), Rebuff (prompt injection defense), and Huntr (the world's first AI/ML bug bounty platform with 17,000+ security researchers). These acquisitions filled out a comprehensive three-product suite covering the full AI security lifecycle. Palo Alto Networks announced its acquisition of Protect AI in April 2025 at RSA Conference, with sources valuing the deal at over $500 million; the transaction closed in July 2025.
Protect AI's technology is built around three flagship products: Guardian (AI model scanning across 35+ formats, continuously scanning over 1.5 million Hugging Face models), Recon (automated LLM red teaming with 450+ attack vectors), and Layer (runtime security for AI applications, tracking conversation flows, tool calls, and multi-turn attacks with 27 turnkey policies). The platform's intelligence advantage comes from the Huntr community — 17,000+ security researchers focused exclusively on AI/ML vulnerabilities — giving Protect AI first-party threat research that no competitor can replicate at the same scale. This research engine, now inside Palo Alto Networks, powers Prisma AIRS 2.0's model scanning and red teaming capabilities.
Protect AI matters because it built the technical foundations of the AI security category before most of the industry acknowledged the category existed. While others were repackaging existing security tools with AI branding, Protect AI was architecting model-native defenses: scanning PyTorch serialization vulnerabilities, detecting architectural backdoors in neural networks, and simulating adversarial attacks against LLMs in production. The company's acquisition by Palo Alto Networks for over $500 million — in just three years from founding, with only $108.5 million raised — is one of the most capital-efficient security exits in recent history and validates that model-level AI security is a genuine enterprise requirement, not a niche concern. For CISOs and AI teams evaluating AI security tools today, Protect AI's technology (now inside Prisma AIRS 2.0) remains the most technically sophisticated AI security stack commercially available.
At acquisition, Protect AI was the most technically comprehensive pure-play AI security vendor in the market, with no direct equivalent competitor offering the same breadth of model scanning, red teaming, and runtime protection in a single integrated platform. Competitors included HiddenLayer (model security, backed by M12 and others), Robust Intelligence (acquired by Cisco in 2024), and AI security features from CNAPP vendors like Wiz and Palo Alto Networks itself. Protect AI's acquisition by PANW effectively ended its independent competitive trajectory, but its technology is now the engine of the industry's most significant AI security platform. The strategic consequence for the market is consolidation: Protect AI's independent existence validated the category; its absorption into PANW validates that AI security will be won by platforms, not point solutions.
206 companies across 10 categories — the most comprehensive AI security company tracker.
Browse All Companies →