HiddenLayer raises $100 million in Series B led by Delta-v Capital

HiddenLayer, the Austin-based startup that builds defenses for models, agents and workflows against adversarial attacks and malicious code injection, closed a $100 million Series B round led by Delta-v Capital. The round came as the company reported a 10-fold increase in annual recurring revenue (ARR) over the past year. CEO and co-founder Chris Sestito declined to give an exact figure, but said the ARR is now in “tens of millions of dollars” and that more than 90% of the growth came from new customers signed in the last twelve months.

The market is maturing rapidly. Three years ago, when the company raised $50 million in a Series A round, the central question was whether AI threats would ever materialize at a scale that justified a real market. Today Gartner estimates that organizations will spend $2.83 billion this year on AI security products, an 83% jump from 2025, with spending expected to reach $4.78 billion next year. Large firms are no longer waiting for headlines about agent misuse; the risk that agents escape control in production environments is enough to drive budgets.

The biggest customers come from financial services and technology firms building AI products, alongside contracts with the U.S. Department of Defense and the intelligence community. One customer is described as a “leading advanced model provider with more than 700 million weekly users,” a description that fits OpenAI or Anthropic. The current round also included Ten Eleven Ventures, Morgan Stanley, Microsoft’s M12, and Booz Allen Hamilton.

Sestito explains that the core technology—runtime protection, detection, attack simulation and supply-chain security—remains relevant as the industry moves from classic ML to GenAI and now to agentic workflows. “Inference remains inference,” he said, comparing the need for runtime security to traditional EDR, but dedicated to AI. The main shift has been expanding coverage to include prompt injection, agent manipulation and malicious tool use.

The company scans about 50 different AI file frameworks to verify that open-weight models are truly what they claim to be. Sestito cited cases of “models hidden inside models,” files that masquerade as one thing while exhibiting different behavior. The new capital will primarily fund sales and distribution, with continued investment in engineering and research, and plans to expand into Europe and the broader EMEA region.

In the background, cyber giants such as Cisco, Palo Alto Networks and Check Point prefer to acquire technology like this rather than build it themselves, while competitors such as Noma and Zenity have already raised tens of millions. With a cash pile of $100 million and this growth trajectory, HiddenLayer positions itself as a natural acquisition target or a potential IPO candidate if market conditions allow.