HiddenLayer Raises $100M for AI Runtime Security

HiddenLayer raised a $100M Series B on September 2, 2026, to expand an enterprise security platform built for agentic, generative, and predictive AI. Delta-v Capital led the financing, with Ten Eleven Ventures, Morgan Stanley, Microsoft’s M12, and Booz Allen Ventures participating.

The Austin company plans to deepen Agentic Runtime Security and Agent Harness Security, products designed to monitor and control AI systems while they take actions across tools, code, data, and infrastructure. The round matters because enterprise AI security is moving beyond scanning models and filtering prompts. Autonomous agents can now execute multi-step work, which makes the behavior between an instruction and an outcome part of the security boundary.

What HiddenLayer Announced

HiddenLayer’s $100M Series B announcement says the capital will support continued investment in its enterprise platform, particularly runtime protection for autonomous agents and coding systems. The company did not disclose a valuation, ownership terms, individual investor allocations, or whether any portion of the transaction involved secondary shares.

The round was led by Delta-v Capital. Ten Eleven Ventures, Morgan Stanley, M12, and Booz Allen Ventures also participated. Ten Eleven previously led HiddenLayer’s $6M Seed round in 2022, while M12 and Moore Strategic Ventures co-led a $50M Series A in September 2023. Those disclosed rounds imply at least $156M in cumulative announced funding, although HiddenLayer did not state an updated lifetime total in the Series B release.

HiddenLayer says annual recurring revenue grew more than 10x during the prior year and that it signed more than 50 new platform customers. In an interview with TechCrunch, co-founder and CEO Chris Sestito said ARR had reached the tens of millions of dollars and that more than 90% of the annual growth came from new customers. The exact ARR figure, contract values, customer retention, and profitability were not disclosed, so those metrics remain company-reported traction rather than independently audited financial results.

Why AI Agents Change the Security Boundary

Traditional software usually exposes a relatively clear sequence: a user makes a request, an application processes it, and controls inspect the input, identity, code, endpoint, network, and resulting data movement. An AI agent can stretch that sequence across a plan, retrieved context, model calls, tools, APIs, credentials, files, and actions taken on the user’s behalf.

The dangerous instruction may therefore arrive after the initial prompt. It can sit inside a webpage the agent retrieves, a package it installs, a tool description it trusts, or data passed from another system. Each individual permission can be valid while the combined behavior becomes unsafe. That makes runtime context important: security teams need to understand what the agent is trying to do, which tools it is using, and whether the current action fits the task and policy.

HiddenLayer’s AI Runtime Security documentation describes real-time monitoring of model inputs and outputs, with the ability to detect, redact, or block content. Its agentic implementation reconstructs the many calls inside a session so investigators can inspect the sequence as one operating record. The company says this approach can identify prompt injection, malicious tool calls, data exfiltration, and unauthorized actions before the workflow completes.

That capability does not replace endpoint, identity, cloud, application, or data-security controls. It addresses a different layer: the behavior of an AI system that has already been given legitimate access and is deciding how to use it.

The Company Behind the Round

HiddenLayer was founded in Austin in March 2022 by Chris “Tito” Sestito, Tanner Burns, and Jim Ballard. The company’s current leadership page lists Sestito as chairman, CEO, and co-founder; Burns as co-founder and chief scientist; Ballard as co-founder and CIO; and Jacob Rideout as CTO.

The founding story began with adversarial-AI research and an attack the team encountered while working at Cylance. HiddenLayer emerged from stealth with a platform focused on defending machine-learning models without requiring access to raw model data or changing the model itself. The company’s scope has since expanded from model detection and response into AI discovery, supply-chain security, automated attack simulation, and runtime protection across generative and agentic systems.

Its platform documentation now presents four connected modules: discovery, supply-chain security, attack simulation, and runtime security. The company says its research team holds 39 granted patents and 65 pending patents and contributes to initiatives involving CISA’s Joint Cyber Defense Collaborative, MITRE, NIST, OWASP, and OpenSSF. Patent counts, customer metrics, and unnamed-customer descriptions are claims from HiddenLayer’s release and should be read with that attribution.

What the Capital Makes Possible

The clearest product priority is securing autonomous execution. HiddenLayer introduced Agent Harness Security in August 2026 for coding agents that can read source code, run commands, modify files, and interact with developer systems. The product integrates with an agent’s hook surface to inspect actions during execution and apply controls around prompt injection, sensitive-data exposure, and unsafe commands.

The Series B also supports engineering, research, sales, channel development, and international expansion, beginning with Europe and the wider EMEA region. HiddenLayer recently appointed Mike Gesnaldo as CRO and says additional hiring is planned. The company has not disclosed headcount targets, a timetable for the expansion, or how the $100M will be allocated among product, research, and go-to-market functions.

The investor logic is visible in the transition from design-time assurance to operating-time defense. As companies move agents from demonstrations into software development, customer service, finance, security, and other workflows, the cost of an unsafe action moves closer to production. A monitoring system that only evaluates the first prompt or final answer can miss the sequence where risk actually developed.

What HiddenLayer Still Has to Prove

The AI-security market is attracting large incumbents and well-funded specialists. Cloud providers, model companies, application-security vendors, and cybersecurity platforms are all extending controls toward AI workloads. HiddenLayer must show that a purpose-built, independent layer can remain valuable as native guardrails improve and established platforms bundle adjacent capabilities.

Enterprise buyers will judge more than detection breadth. They will want low false-positive rates, explainable evidence, manageable latency, policy controls, integration with existing security operations, and response that does not stop legitimate work. HiddenLayer’s growth claims suggest the need is moving into real budgets, but customer names, renewal data, deployment depth, and independent outcome measurements remain limited in the public record.

The company’s advantage may come from following an agent across the seams that separate models, tools, code, and infrastructure. Those seams are where one authorized action becomes a risky sequence. The Series B gives HiddenLayer room to turn its research lead into a durable operating layer while autonomous systems keep expanding the set of things software is allowed to do.