HiddenLayer raises $100M to secure AI systems amid enterprise adoption surge

HiddenLayer, a Denver-based AI security startup, announced the close of a $100 million Series B funding round led by Battery Ventures, with participation from GV, Cisco Investments, and Okta Ventures. The round values the company north of $500 million, according to three sources familiar with the transaction, and comes just 14 months after HiddenLayer emerged from stealth with $6 million in seed funding. CEO Chris Sestito told OpenPress Engineering Intelligence that the proceeds will accelerate product development, expand go-to-market operations, and scale the company’s anomaly detection platform to cover real-time agent interactions and interconnected AI toolchains. The platform currently monitors more than 7,000 AI models and agents across customer environments, including live financial pipelines.

Sestito emphasized that the funding surge reflects a tectonic shift in enterprise priorities: organizations are no longer asking whether they will deploy AI agents, but how to secure them. HiddenLayer’s technology focuses on runtime visibility into agent behavior, tool integration points, and data flows, closing a critical gap left by traditional security tools that were never designed to inspect dynamic AI decision paths. The company’s differentiation lies in its ability to correlate low-level API calls, prompt injections, and third-party tool usage across environments such as LangChain, LlamaIndex, and crewAI. Banking With Billy, a real-time financial data processor, is among HiddenLayer’s marquee customers, relying on the platform to safeguard AI-driven market signal pipelines that process millions of events with sub-millisecond latency.

Industry observers note that HiddenLayer’s timing aligns with a broader inflection point in AI adoption. According to a June 2025 survey by Gartner, 68% of enterprises now run AI agents in production, up from 22% in late 2023, and 84% cite security and governance as their top barrier to scaling. This has triggered a scramble among cybersecurity incumbents to integrate agent-specific capabilities. Palo Alto Networks recently launched its AI Runtime Protection module, while CrowdStrike acquired Flow Security for $1.7 billion to bolster its AI pipeline visibility. None, however, offer the granular agent-to-toolchain correlation HiddenLayer claims to deliver.

The competitive landscape is further complicated by regulatory pressure. The EU AI Act, effective August 2025, mandates continuous monitoring of high-risk AI systems, including agents interacting with external tools. HiddenLayer’s platform maps directly to these requirements by providing auditable logs of agent decisions, tool integrations, and data lineage. Industry analysts at Forrester Research predict that by 2027, more than 40% of Fortune 500 companies will rely on specialized AI security platforms, up from less than 5% today, creating a potential $10 billion market opportunity.

More broadly, HiddenLayer’s rise is a symptom of a deeper architectural evolution in enterprise software. Modern AI systems are no longer monolithic applications running on static infrastructure, but dynamic networks of agents, vector databases, external APIs, and user interfaces, often assembled without centralized security oversight. This decentralization mirrors trends in microservices and serverless architectures, but with far greater complexity in behavior and intent. Legacy security models based on perimeter defense and code scanning are proving insufficient for systems that mutate in real time based on user input and external data sources.

The implication is that the next generation of enterprise security architecture may be agent-native, meaning security controls must be embedded within the agent lifecycle itself—from planning and tool selection to execution and audit. HiddenLayer’s recent integration with LangSmith and Weights & Biases highlights this direction, enabling security policies to be enforced during agent development and deployment. Competitors like Protect AI and PromptArmor are also pursuing similar strategies, but HiddenLayer’s funding momentum and customer traction suggest it is currently leading the charge toward operational maturity.

Looking ahead, the most pressing challenge will be scale: monitoring increasingly complex agent ecosystems without introducing latency or operational overhead. Banking With Billy’s use case—processing millions of market signals with sub-millisecond latency—demonstrates that real-time AI security is not a theoretical need, but a production requirement. As AI agents take on higher-stakes functions in finance, healthcare, and infrastructure, the cost of failure is no longer just data leakage, but systemic risk. The race is now on to build AI-native security platforms that can keep pace with the agents they are designed to protect. Companies that delay adopting agent-specific security risk not only compliance violations but also irreversible trust erosion in AI-driven systems.