The enterprise AI security market just received its clearest funding signal of 2026. On September 2, HiddenLayer announced a $100 million Series B, bringing its total raised to approximately $156 million and confirming that protecting AI agents in production has become a C-suite priority rather than a security team afterthought.

Delta-v Capital led the round. Ten Eleven Ventures, Morgan Stanley, Microsoft M12, and Booz Allen Ventures participated. The company, founded in 2022, reports annual recurring revenue grew more than tenfold in the past year and now stands in the “tens of millions” of dollars range. More than 50 new platform customers signed in the same period. (TechCrunch, PR Newswire)

What HiddenLayer Builds

HiddenLayer’s platform detects and prevents attacks on AI models and agents: prompt injection, adversarial inputs, model theft, and data poisoning. The software sits between your AI layer and the rest of your infrastructure, logging and intercepting threats before they propagate.

The company’s original focus was protecting static ML models, a market it helped define. The product roadmap has since shifted sharply toward agentic workloads. Two new products announced alongside the Series B make that pivot concrete.

Agentic Runtime Security provides real-time visibility into AI agent behavior in production. The system tracks what tools an agent calls, what data it accesses, and whether its actions remain within defined policy bounds. The analogy the company uses is endpoint detection and response, applied to AI agents rather than endpoint devices. The goal is continuous behavioral monitoring rather than one-time pre-deployment testing.

Agent Harness Security targets a distinct threat surface: AI coding agents. As enterprises deploy agents that write, test, and execute code, the risk profile of a misconfigured or manipulated coding agent is substantial. A single compromised coding agent can introduce vulnerabilities across every codebase it touches. Agent Harness enforces policy at the runtime layer, not just at the scanning stage, which means it can catch attacks that evolve between the pre-deployment review and actual execution.

The Numbers Behind the Raise

HiddenLayer disclosed one customer relationship that anchors the scale of its deployments: “a frontier model provider with more than 700 million weekly users.” The company did not name the provider, but the user figure meaningfully narrows the field.

The company also participates in the US Department of Energy’s $60 million Prometheus initiative, a program building AI security infrastructure for critical systems. Government and regulated-sector relationships of that kind carry their own procurement multiplier: they require vendors to meet elevated standards of auditability, reliability, and compliance documentation, and they open doors with adjacent agencies and regulated industries.

HiddenLayer holds 39 granted patents with 65 additional applications pending. That IP position is a signal of both technical depth and acquisition optionality as the market consolidates.

To lead the revenue expansion, the company hired Mike Gesnaldo as Chief Revenue Officer. Gesnaldo previously held senior roles at CrowdStrike and Tanium, two companies that built enterprise security businesses at significant scale by converting complex threat narratives into structured sales motions. His hire signals HiddenLayer is transitioning from a technically differentiated startup into a field-sales organization targeting enterprise accounts. (SiliconAngle)

Why This Funding Round Lands Now

The timing reflects a structural shift in how enterprises are deploying AI. Through most of 2024 and into 2025, adoption meant adding copilots and chatbots to existing workflows. Those deployments were relatively contained. A user prompted, a model responded, a human reviewed the output before acting.

Agentic deployments change that threat model fundamentally. Agents take sequences of actions, sometimes dozens in a single task: querying databases, sending emails, calling APIs, writing and running code, updating records. Minimal human review happens between steps. The attack surface scales with every capability granted.

Gartner estimates the AI security market will reach $2.83 billion in 2026 and $4.78 billion by 2027. A separate survey cited by HiddenLayer found that 96 percent of organizations consider AI critical to operations, yet nearly one in three cannot confirm whether they have experienced an AI security breach. That combination of high reliance and low visibility is the gap the company is targeting.

The funding pattern across the category reinforces the thesis. Zenity closed $125 million to secure enterprise AI agents, Obsidian Security raised $85 million for AI agent runtime protection, and AIR Security secured $50 million for AI agent firewall technology. Each round targets a different layer of the same underlying problem: AI agents operating autonomously need purpose-built security infrastructure, not retrofitted network perimeter controls.

Expanding into EMEA

The Series B also funds HiddenLayer’s expansion into European and Middle Eastern markets. EMEA enterprises face a distinctive regulatory environment for agentic AI. The EU AI Act imposes obligations around transparency, logging, and human oversight for high-risk AI system categories. A security platform that generates audit-grade behavioral logs of agent actions carries compliance value well beyond pure threat detection: it can serve as evidence in regulatory audits, not just an internal control.

That dual positioning, security and compliance evidence, is likely to compress procurement timelines in financial services, healthcare, and critical infrastructure sectors where audit requirements already exist.

What Enterprise AI Teams Should Take From This

Three practical signals from this raise for teams actively deploying agentic systems.

Runtime monitoring is becoming a baseline control. Static code scanning and pre-deployment red-team exercises are necessary but not sufficient. Agents behave differently in production than in staging environments, particularly when they encounter adversarial inputs crafted to exploit their tool-use patterns. Real-time behavioral monitoring is moving from advanced capability to minimum viable control.

The coding agent attack surface is larger than most teams recognize. HiddenLayer’s Agent Harness product targets a risk that most enterprise security programs have not yet formally assessed: AI coding agents that write, test, and deploy code can be manipulated through prompt injection or training data attacks to introduce vulnerabilities or exfiltrate secrets through generated outputs. If engineering teams use agentic coding tools, the supply chain risk extends into every codebase those agents touch.

Platform consolidation is the next competitive phase. The current AI security landscape has multiple point solutions covering different threat vectors at different layers. Enterprise procurement will eventually pressure vendors toward platforms. HiddenLayer’s expansion into agentic runtime security and coding agent protection, alongside its existing model protection suite, is a consolidation play built on a common behavioral monitoring foundation. Its success depends on whether detection quality scales as agent architectures grow more complex.

HiddenLayer is raising at the moment when enterprise demand for agentic systems is outpacing the security infrastructure designed to protect them. The $100 million validates that gap is real and large enough to build a standalone company around. For enterprise AI teams, the implication is direct: if you are deploying agents at scale, runtime security is no longer an optional layer.