HiddenLayer nabs $100M as enterprises rush to secure their AI deployments

Three years ago HiddenLayer's $50 million Series A came with an open question: would attacks on models show up often enough to be a business? TechCrunch's Kyle Wiggers noted at the time how few public examples existed. That question is closed enough to raise a $100 million Series B led by Delta-v Capital, with Ten Eleven, Morgan Stanley, Microsoft's M12, and Booz Allen in the round.

Gartner estimates companies will spend $2.83 billion this year securing AI tools, 83 percent more than 2025, and nearly $4.78 billion next year. CEO Chris Sestito says HiddenLayer's ARR grew more than 10x in a year to the 'tens of millions,' over 90 percent from new customers. Finance and big tech are the largest verticals; it also has Defense and intelligence contracts, and a customer described as a frontier lab with more than 700 million weekly users.

HiddenLayer did not invent a new category so much as stretch an old one. Sestito's line is that inference is still inference: the same discovery, runtime protection, attack simulation, and supply-chain scanning now have to cover prompt injection, agent manipulation, and malicious tool use. He likens runtime to EDR for AI. The new work is parsing about 50 model-file frameworks so an open-weight file is the model it claims to be, not a hidden model inside a model.

That is a good product story and a crowded one. Noma and Zenity have each raised more than $100 million in adjacent AI-security ground. Cisco, Palo Alto, and Check Point would rather buy this layer than rebuild it. Sestito admits Microsoft, OpenAI, and AWS could eventually bundle pieces of what HiddenLayer sells. His bet is that the platforms will stop at discovery, identity, and policy, and leave runtime guts to a specialist.

Signal Desk's read: the 10x ARR and DoD logos buy HiddenLayer a window, not a moat. The $100 million is going into sales, Europe, and more engineering. If a platform ships 'agent EDR' as a checkbox in the next two enterprise cycles, that window closes. The honest test is whether HiddenLayer can stay the system of record for model-file integrity after the first major open-weight poisoning incident — or whether that incident is what gets it acquired.

Austin-based HiddenLayer still sells the 2023 shape: discovery, runtime, simulation, supply chain. The Series B is a distribution raise on top of a scope expansion from classical ML to gen AI to agents. Gartner's spend curve is the demand slide every vendor in this room is using.