Three years ago, when HiddenLayer raised its $50 million Series A, the AI security startup faced a fair question: would attacks on AI systems ever happen at a scale that justified a real market? The answer, apparently, is yes. Gartner now estimates companies will spend $2.83 billion this year on products to secure AI tools — an 83% jump over 2025 — and nearly $4.78 billion next year. HiddenLayer has ridden that wave hard. CEO Chris Sestito tells TechCrunch that annual recurring revenue grew more than 10x over the past year, landing in the “tens of millions” of dollars, with over 90% of that growth coming from new customers.
That momentum convinced investors to back a $100 million Series B led by Delta-v Capital, with participation from Ten Eleven Ventures, Morgan Stanley, Microsoft’s M12, Booz Allen Hamilton, and others. The Austin-based company protects AI models, agents, and workflows from adversarial attacks, vulnerabilities, and malicious code injections. Its customer base skews toward financial services and large tech companies building AI products, along with contracts with the Department of Defense and the intelligence community. One customer, described as a “leading frontier model provider” with “more than 700 million weekly users,” sounds a lot like OpenAI or Anthropic.
Sestito is careful to frame the company’s evolution as expansion rather than a pivot. “Inference is still inference,” he told TechCrunch. “Whether it’s on a traditional machine learning model, whether it’s GenAI, whether it’s an agentic work stream, a lot of our technology still applied.” The biggest change has been extending existing products — discovery, runtime protection, attack simulation, and supply chain security — to address prompt injection, agent manipulation, and malicious tool use. Sestito likens runtime security for AI to traditional endpoint detection and response, or EDR, which has been a staple of enterprise security for years.
One area HiddenLayer is watching closely: open-source and open-weight models. The company parses and scans about 50 different AI file frameworks to catch models that are not what they claim to be — including hidden models embedded inside other models. The new capital will go toward sales and distribution, with plans to expand into Europe and EMEA. But the competitive landscape is getting crowded. Cisco, Palo Alto Networks, and Check Point often prefer to acquire this kind of technology rather than build it, while startups like Noma and Zenity have each raised over $100 million to tackle adjacent problems. Sestito acknowledged that parts of his company’s offerings could eventually be bundled into platforms from Microsoft, OpenAI, or AWS, but he expects those giants to focus on governance features like discovery, identity, and policy controls rather than the deeper security tooling HiddenLayer builds. The startup’s stated ambition is to “scale vertically alongside artificial intelligence” and then expand horizontally into AI-dependent cybersecurity. That’s a bold plan, but the first test is simpler: prove the head start translates into a durable business before the rest of the industry closes the gap.
