Key Takeaways

  • This $100 million Series B reflects rising demand for security controls built specifically for AI models, agents, and workflows.
  • Delta-v Capital led the round, with Ten Eleven Ventures, Morgan Stanley, Microsoft’s M12, and Booz Allen Hamilton participating.
  • The investment arrives as enterprises confront model poisoning, adversarial manipulation, and AI supply-chain risks that conventional security products may not fully address.

Investors previously debated whether AI security would become a distinct market or simply another feature inside existing cybersecurity platforms. HiddenLayer's $100 million Series B offers a clear answer: enterprises are moving fast to secure AI models, agents, and workflows against adversarial manipulation and software supply-chain risk.

Delta-v Capital led the round, with participation from Ten Eleven Ventures, Morgan Stanley, Microsoft’s M12, and Booz Allen Hamilton. The mix is notable. It brings together security-focused investors, financial institutions, a major technology platform, and a large government and enterprise consultancy, suggesting that demand for AI defense is appearing across several buyer groups.

HiddenLayer focuses on protecting AI models, agents, and workflows from threats such as adversarial manipulation and software supply-chain compromise. Those risks can emerge at multiple stages. A model might be poisoned during training, altered while stored in a repository, or manipulated through malicious input after deployment. Agentic systems widen the exposure because they can access data, call external tools, and initiate actions.

That changes the security equation. Traditional endpoint and network products can still protect the infrastructure surrounding an AI application, but they may have limited visibility into model behavior, prompt-based attacks, or subtle changes in outputs. Dedicated AI security layers are intended to monitor those interactions and identify activity that looks abnormal or potentially hostile.

Securing AI and using AI for security are related, but they are not the same market. The latter applies machine learning to areas such as fraud detection, threat hunting, and anomaly identification. Grand View Research estimated the global AI in cybersecurity market at $25.35 billion in 2024 and forecasts it will reach $93.75 billion by 2030, representing a 24.4% compound annual growth rate.

Securing AI turns that relationship around. It treats models and AI-enabled applications as assets requiring their own testing, access controls, monitoring, and incident-response processes. Gartner’s 2026 forecast projects worldwide information security spending of $248.9 billion in 2026 and $372.6 billion by 2030. Within that total, spending on securing AI is expected to rise from $15.6 billion in 2025 to $37.6 billion in 2030.

The threat environment helps explain the urgency. ENISA’s Threat Landscape 2025 examined 4,875 incidents recorded between July 2024 and June 2025. Its findings highlighted the growing use of AI by attackers and defenders, including automated phishing and model-poisoning techniques.

A practical challenge remains: ownership. Should AI security sit with the chief information security officer, the data science organization, or the application team deploying a model? In many enterprises, responsibility is still distributed. That can leave gaps between model development, cloud security, and production monitoring.

Frameworks such as the NIST AI Risk Management Framework, 2023 and the NIST Cybersecurity Framework (CSF) 2.0, 2024 offer useful reference points. Both emphasize governance, secure design, and ongoing monitoring rather than treating security as a final checkpoint before deployment. For businesses, that can mean cataloging models, assessing third-party components, and defining how suspicious AI behavior will be investigated.

The company is not entering an empty field. AI-security specialists such as Huskeys Security are working at the edge, while Microsoft and Booz Allen Hamilton are building, buying, and funding capabilities across broader AI pipelines. Competition will likely center on integration, visibility, and whether products can protect diverse models without slowing production workloads.

Still, the Series B is more than a startup financing milestone. It indicates that buyers and investors increasingly expect AI protection to become an operational layer within enterprise technology stacks. As autonomous agents gain access to sensitive data and business systems, the market will be shaped by a straightforward question: can security teams observe and control what those systems are doing before an unusual interaction becomes a material incident?