Enterprise AI agents can now read data, call APIs and change records, and a new $125M round for Zenity signals how seriously security teams are taking that shift. Norwest led the Series C, with SoftBank Vision Fund 2, Qumra Capital, Hitachi Ventures and LG Technology Ventures among the new investors. Vertex Ventures, Third Point Ventures, DTCP and Intel Capital returned, and total disclosed funding now tops $180M.
The company’s platform watches what agents actually do: which tools they connect to, what permissions they hold, what they remember and how they behave at runtime. Its reach includes SaaS copilots such as Microsoft 365 Copilot and Salesforce Agentforce, custom agents running on AWS Bedrock, Azure AI Foundry or Google Vertex AI, and coding assistants installed on employee devices.
Zenity argues that the first wave of generative AI was mostly chat, while agents are different because they act. A prompt filter alone cannot catch a harmful outcome built from a chain of individually harmless steps, so its detection layer examines the execution path and can block an action before it happens.
Research from Zenity Labs has shown zero-click and one-click attacks where instructions hidden in emails, documents or support tickets push agents into leaking data or abusing connected tools. The company frames the problem as an agent-layer issue, separate from the models underneath.
The raise lands as enterprises prepare for what Zenity calls the era of one billion AI agents, and a valuation was not disclosed.
