Why We Invested in HiddenLayer

Today, we are pleased to announce Delta-v's investment in HiddenLayer, an Austin-based AI security company purpose-built to protect AI models, applications, and agents. Delta-v is leading the company's $100 million Series B, with meaningful participation from Ten Eleven Ventures and a broader syndicate of strategic and financial investors. The new capital will support continued investment in HiddenLayer's platform, including its Agentic Runtime Security capabilities and its work securing autonomous coding agents.

Over the past several years, enterprises have moved quickly from experimenting with AI to deploying it inside customer-facing products, software development workflows, and core business operations. That transition has created a new security problem. Traditional cybersecurity tools were built to protect networks, endpoints, applications, identities, and data, but they were not designed to understand how an AI model behaves, determine whether it has been tampered with, or recognize when an autonomous agent is being manipulated into taking unintended actions.

That gap is what led Delta-v to canvas the landscape of AI security startups starting in 2024, well before the category had reached its current level of commercial activity. As we studied the market, we came to believe that the AI security category would favor new companies built around the underlying AI architecture from the beginning.

Our work in AI security grew out of Delta-v's broader focus on cybersecurity and infrastructure software, where we have seen major shifts in technology create new security requirements and, in turn, new categories of security companies.

As the many in the industry have noted, the move to the cloud offers the clearest comparison. It did not simply change where applications were hosted, but also how infrastructure was provisioned, how data moved, and how software was built and operated. Those changes required a different security architecture and created room for products designed specifically around cloud environments. We believe AI is creating a similar architectural shift.

Cloud security products can see workloads, configurations, identities, and network activity, but they generally cannot see what is happening inside a model or agent. They cannot determine how a model is responding to a prompt, whether a model artifact has been compromised, or what an agent is attempting to do through the tools and permissions available to it.

That led us to a simple view: AI security would require its own purpose-built architecture, much as cloud security did during the last major infrastructure shift.

An AI system can be compromised without a conventional network breach. Attackers might manipulate a model's inputs to change its behavior, attempt to reconstruct or steal a proprietary model, introduce a compromised model through the AI supply chain, or cause an agent to use its permissions in a way the enterprise did not intend.

These risks become more consequential as AI use cases evolve from producing information to taking action. Security teams therefore need visibility into the AI stack itself: what it receives, what it produces, how it behaves, and what actions it attempts to take. Existing tools were generally not designed to provide that visibility.

As enterprises place more models, applications, and agents into production, they will have more assets to discover, test, monitor, and protect. We do not view AI security as a temporary spending theme, but rather an emerging layer of the enterprise security architecture.

First, we believe AI security is becoming a requirement as enterprises put AI into production and give models and agents access to sensitive data, software, and business systems.

Second, we believe the category will favor purpose-built platforms because the relevant risks sit inside model behavior, inference pipelines, AI supply chains, and agent activity, which makes them difficult to address with existing security products.

Third, we believe HiddenLayer is uniquely positioned to serve this market. It's a research-led company; HiddenLayer has already identified dozens of vulnerabilities across models and the surrounding AI infrastructure. As an early leader, its production deployments are learning how these systems are attacked and misused in practice. The combination of cutting-edge research and real-world data should continue to deliver value for customers as the company scales.

HiddenLayer was founded in 2022 by a team that had already encountered the problem firsthand while responding to an early real-world adversarial AI attack. Rather than begin with a conventional security product and extend it toward AI, the team built the platform around the behavior and structure of AI systems. HiddenLayer helps enterprises identify the AI assets operating across their environments, inspect models before they are deployed, test applications for vulnerabilities, and monitor models and agents while they are running.

Over the past year, the company grew annual recurring revenue by more than 10x and signed more than 50 new platform customers, including some of the largest names in securities brokerage, banking, insurance, accounting, government, technology, IT services, pharmaceuticals, and airlines.

HiddenLayer's research team has uncovered dozens of vulnerabilities across the AI ecosystem, including issues affecting foundation models and the tooling and infrastructure around them. That work feeds directly into the product, while production deployments give the company additional insight into how attacks behave outside a lab environment.

The platform has also been tested in environments where the security bar is unusually high. HiddenLayer is deployed across multiple systems within the U.S. Department of Defense and intelligence community, and it supports a leading frontier model provider in securing more than 700 million weekly users.

HiddenLayer sits at the intersection of two areas Delta-v has followed for years: the infrastructure needed to deploy AI at scale and the security platforms enterprises depend on to operate safely.

Chris Sestito and the HiddenLayer team identified the need for AI security before most enterprises had meaningful AI systems in production. They built for the problem early, remained focused as the category developed, and have earned the trust of customers operating some of the most demanding AI environments in the market.

We believe AI security will become a durable part of the enterprise technology stack, and that HiddenLayer has the architecture, research depth, and customer validation to become one of the defining companies in the category. With our infrastructure software team based in Dallas, we are very pleased to partner with a dynamic company being built in Austin.

The information contained herein is for informational purposes only and does not constitute an offer to sell or a solicitation of an offer to buy any securities or investment products. Certain statements may be forward-looking in nature and are subject to risks and uncertainties; actual results may differ materially. References to customer experiences are anecdotal and should not be construed as endorsements or investment advice. Past performance is not indicative of future results.