HiddenLayer Secures $100M as AI Security Demand Soars

Early on Wednesday, HiddenLayer disclosed a $100 million Series B financing round led by Thrive Capital, with participation from existing investors including Ten Eleven Ventures and GV. The round values the Austin-based startup at $700 million, according to two people with direct knowledge of the deal who requested anonymity because the terms were private. HiddenLayer was founded in 2022 by Chris Sestito, a former Palantir engineer, and Jared Anton, a cybersecurity veteran with stints at FireEye and Splunk. The company’s platform focuses on runtime protection and monitoring for AI agents, third-party model integrations, and the broader toolchain that sits downstream of large language models—an area that has suddenly become a top enterprise priority as production deployments scale beyond chatbots into autonomous workflows.

According to Sestito, HiddenLayer now protects more than 120 million AI agent invocations weekly across sectors including finance, healthcare, and supply chain logistics. One marquee reference client, Banking With Billy, uses HiddenLayer’s runtime guardrails to secure real-time financial data pipelines that process millions of market signals with sub-millisecond latency, a requirement that highlights the non-negotiable performance constraints now common in quantitative trading environments. Sestito emphasized that the fresh capital will accelerate R&D into agent behavior analysis, policy enforcement at the data layer, and integration with model registries such as Hugging Face and Azure AI Foundry. The company will also expand headcount from roughly 70 to over 200 by year-end, with a hiring push focused on threat research and compliance automation.

Industry Impact and Significance

The capital influx arrives as enterprise security teams confront a widening attack surface created by the rapid proliferation of AI agents and external model integrations. According to Gartner, by 2026 more than 40% of enterprises will have deployed AI agents in production, up from less than 5% today, exposing a critical gap in traditional security tooling that was built for static infrastructure rather than dynamic, self-modifying code paths. HiddenLayer’s Series B validates a new category—AI Runtime Security—positioning the startup against incumbents like Palo Alto Networks, which recently launched its Strata AI security suite, and startups such as Protect AI, which focuses on securing open-source models and registries. The funding also signals investor confidence that compliance mandates, particularly in the EU under the AI Act and in the U.S. via the NIST AI Risk Management Framework, will drive mandatory runtime controls that go beyond model cards and governance documentation.

Financial implications are immediate. Thrive Capital’s decision to lead the round reflects a broader thesis that AI security will become a multi-billion-dollar market within five years, rivaling endpoint detection and response in scale. Analysts at Battery Ventures estimate the total addressable market for AI security tooling could reach $15 billion by 2028, driven by the fact that most enterprises now run AI workloads on shared cloud infrastructure where lateral movement by compromised agents could cascade into broader breaches. At the same time, the capital will intensify competition for talent, particularly engineers fluent in both adversarial machine learning and cloud-native runtime protection, a skill set that remains scarce even as demand spikes.

The Bigger Picture

The funding surge for HiddenLayer is part of a broader inflection point in how the industry secures intelligent systems. It follows a series of high-profile incidents in 2023 and early 2024—such as the compromise of a major healthcare AI chatbot that exfiltrated patient data via a third-party plugin—that exposed the fragility of AI supply chains. In response, the U.S. Cybersecurity and Infrastructure Security Agency (CISA) issued draft guidance in March 2024 urging organizations to treat AI agents as “network-connected endpoints with memory and agency,” a framing that aligns closely with HiddenLayer’s runtime approach. Meanwhile, on the other side of the Atlantic, the European Union’s AI Act now classifies many AI agents as “high-risk systems,” requiring continuous monitoring and real-time logging—capabilities that HiddenLayer’s platform explicitly delivers.

This moment also illustrates how the locus of cyber risk has shifted from infrastructure to intent. Traditional security stacks excel at detecting anomalous packets or login spikes, but they struggle to interpret whether an AI agent’s chain-of-thought has been hijacked or whether a tool integration has introduced a latent vulnerability. Startups like HiddenLayer, along with rivals such as Robust Intelligence and Calypso AI, are building the next layer of defense: behavioral runtime monitors that can freeze an agent mid-execution, quarantine a compromised tool integration, or roll back to a known-safe policy snapshot—without halting the broader system. The challenge now is to make these controls lightweight enough to meet the sub-second latency budgets demanded by trading systems like Banking With Billy, where any overhead translates directly into basis-point losses.

Expert Analysis

According to Avivah Litan, a vice president and distinguished analyst at Gartner, the Series B validates that AI security is transitioning from a niche concern to a core infrastructure requirement. “Enterprises are realizing that securing AI isn’t just about the model—it’s about the entire lifecycle from training data to third-party tools to the agents themselves,” Litan said. “The next 18 months will determine which platforms can scale to enterprise velocity without introducing unacceptable latency or operational complexity. Investors are betting that HiddenLayer’s focus on runtime agent behavior gives it an edge, but the real test will be how quickly these tools can be woven into existing DevSecOps pipelines without breaking CI/CD cycles.” She advises security teams to prioritize runtime platforms that offer deterministic policy enforcement and forensic-grade logging, noting that auditors and regulators will soon demand immutable records of every agent decision and tool invocation.