Onyx Security raises $113M to govern enterprise AI agents

Onyx Security announced on Wednesday it raised $113 million in a Series B round led by Bessemer Venture Partners. Other participants included Cyberstarts, TCV, Conviction, FirstMark, Vintage, QuantumLight and G Squared. The new funding brings Onyx’s total financing to $153 million since the company was founded two years ago. The startup did not disclose an official valuation; reports put it at about $640 million after the round. Onyx said it will use the capital to develop its proprietary models and scale sales and customer support.

The company offers a centralized platform that detects and regulates advanced AI agents operating across SaaS, cloud and endpoint environments. The platform is designed to monitor agent actions, apply policy controls and record activity when agents interact with sensitive data or critical systems.

Onyx says its technology tracks an AI agent’s decision-making at each step and can intervene in real time to correct unintended or malicious behavior. The product also aims to find unauthorized or “shadow” AI deployments and enforce safeguards against attacks such as prompt injection.

Onyx describes its stack as an operational control layer that applies policy and accountability to agents as they carry out tasks that touch sensitive data, systems and processes. Company leaders have emphasized that the platform uses Onyx’s own models to monitor agent behavior and enable automated interventions across multiple enterprise environments.

Investors in the round have backgrounds in cybersecurity and enterprise software. The funding comes while more organizations test generative models and agentic assistants that can perform multi-step actions across software and cloud services.

Onyx said the Series B follows two years of product development and early customer deployments. The company plans to direct the new funds to product work, model training and expanding market and support operations.

Hila Zigman, general partner at Cyberstarts, commented that “Onyx is building the control layer that makes enterprise AI adoption possible at scale,” and added that her firm expects the security category for agent control to expand over the next decade.