Protect AI raised $108.5M, then Palo Alto Networks completed its acquisition on 22 July 2025. Its model scanning now ships as AI Supply Chain Security inside Prisma AIRS, sold through sales. The open-source LLM Guard library was archived on 9 July 2026.
Protect AI was a Seattle AI and ML security company that raised $108.5M in total, including a $60M Series B led by Evolution Equity Partners in August 2024. The buyer, Palo Alto Networks, closed the deal on 22 July 2025, and the technology now lives in Prisma AIRS rather than a standalone product.
Founded: 2022 · HQ: Seattle, WA, USA · CEO: Ian Swanson (CEO until the acquisition) · Funding: $108.5M raised in total as of August 2024 (Series A 2023, Series B 2024). Acquired by Palo Alto Networks; the buyer did not state a price.
About Protect AI
Protect AI was an AI and machine learning security company headquartered in Seattle, with offices in Berlin and Bangalore, founded by AI leaders from Amazon and Oracle. Ian Swanson was co-founder and CEO. Its thesis was that models moving into production created attack surfaces ordinary security tools did not cover: model files that can run code when loaded, poisoned data pipelines, and prompt injection against LLM applications. The acquisition closed on 22 July 2025 and Swanson moved to Palo Alto Networks as VP, Product for Prisma AIRS. The company no longer sells as a standalone vendor: protectai.com now redirects to Palo Alto Networks' Prisma AIRS pages.
Before the deal, the company described its platform as AI security posture management (AI-SPM): visibility, remediation, control and governance over ML models, LLMs and AI applications. In its August 2024 Series B announcement it said the platform had grown from one product to five and that it had made four acquisitions: Rebuff, Huntr, Laiyer AI and SydeLabs. Laiyer AI was the team behind the open-source LLM Guard toolkit, and SydeLabs added SydeBox, an automated red-teaming tool for GenAI systems. The huntr community, which pays researchers to find vulnerabilities in AI and ML open-source projects, was reported at more than 15,000 members. Palo Alto Networks says the acquisition brings model scanning, posture management, AI red teaming, runtime protection and agent security together inside Prisma AIRS.
Today the model-scanning lineage lives on as AI Supply Chain Security, a Prisma AIRS product that scans AI models and agent skills before deployment for code that executes at load or runtime, architectural backdoors, hidden instructions and license or policy violations. Scans run locally or in the cloud and plug into CLI, API, CI/CD and model-registry workflows. Palo Alto Networks does not publish list prices for it; buyers go through sales. The open-source LLM Guard repository at github.com/protectai/llm-guard was archived by its owners on 9 July 2026 with a notice ending active development and maintenance, so teams that relied on it for prompt and output scanning need a maintained alternative.
Funding is the part most pages get wrong. The company's own Series B release (August 2024) reports a $60M round led by Evolution Equity Partners, a $35M Series A a year earlier, and $108.5M raised in total, with 01 Advisors, StepStone Group, Samsung and returning backers taking part. Palo Alto Networks did not state a purchase price in its completion release, so no price is given here.
Where the alternative wins: if you want a specialist vendor that still ships as an independent product, HiddenLayer is the closest peer in model security and is still sold on its own. If your models live on a public hub, scanning also happens at the registry itself, as with Hugging Face. Protect AI's technology is best judged now as a Prisma AIRS capability rather than as a company you can sign a separate contract with.
Mission
Make AI and machine learning systems safe and secure for enterprise production use by addressing the attack surfaces unique to AI: model tampering, data poisoning, and prompt injection.
Products
- AI Supply Chain Security (Prisma AIRS) (Model and agent-skill scanner): https://www.paloaltonetworks.com/ai-security/ai-supply-chain-security
- LLM Guard (archived) (Open-source LLM security toolkit, no longer maintained): https://github.com/protectai/llm-guard
Links
Frequently Asked Questions
What happened to Protect AI?
The deal closed in July 2025 and Palo Alto Networks absorbed the technology and team into its Prisma AIRS AI security platform. The protectai.com address now forwards to Prisma AIRS pages, and the buyer's completion notice gives no purchase price.
How much funding did Protect AI raise?
By its own count in August 2024, Protect AI had raised $108.5M in total, with a $35M Series A about a year before the Series B. Evolution Equity Partners led the later round, and Salesforce Ventures was among the returning backers.
What does Protect AI's technology do now?
Its model-scanning work continues inside Prisma AIRS as AI Supply Chain Security. The product checks models and skills for code that runs at load time, backdoors, hidden instructions and license violations, either on your own infrastructure or through the cloud. Palo Alto Networks lists no public prices.
Is the LLM Guard open-source library still maintained?
No. The protectai/llm-guard GitHub repository became read-only when its owner archived it on 9 July 2026. The code stays available under the MIT license for existing users, but nobody is fixing bugs or updating the scanners, so new projects should pick a maintained guardrail library.
Who are Protect AI's closest competitors?
HiddenLayer is the nearest independent peer in model security and is listed on Hokai. Protect AI itself no longer sells standalone, so the practical comparison is Prisma AIRS against specialist AI security vendors, plus registry-side scanning on hubs such as Hugging Face. Judge each on model-file scanning depth, agent and skill coverage, and runtime protection.
Top Alternatives
- HiddenLayer: Pick HiddenLayer for an independent model-security vendor; Protect AI's technology now ships inside Palo Alto Networks' Prisma AIRS.