Zenity secures $125 million to focus on AI agents rather than models.

      The majority of funds allocated to AI security have focused on the model and the prompts. However, Zenity has recently raised $125 million by emphasizing a different concern: the actual risk lies in AI agents that operate independently. The Israeli startup aims to protect the actions of these agents once a company integrates them into its systems.

      Norwest spearheaded the Series C funding round, with participation from SoftBank’s Vision Fund 2, Hitachi Ventures, and LG Technology Ventures, as announced by the company. This brings Zenity’s total funding to approximately $185 million, although the company did not disclose its valuation. Notably, the new investors are all companies that are utilizing agents within their own operations.

      Zenity was established in 2021 by two veterans from Unit 8200, Ben Kliger and Michael Bargury, who previously developed security products at Microsoft. The company currently employs over 230 people, with research based in Tel Aviv and a sales team located in New York. According to Zenity, its clientele primarily consists of Fortune 500 and Global 2000 companies in regulated sectors.

      Focusing on securing the agent instead of the model, Zenity’s proposal highlights a shift that the industry is only beginning to recognize. While a chatbot merely responds to queries, an agent takes action. It has the capability to access internal databases, utilize tools, update records, and execute multi-step processes across various systems. This transforms the issue from one of content to one of control.

      An agent can operate as intended and still lead to a security breach. It may have excessive permissions or be susceptible to misleading commands. Zenity monitors the agent layer, including its permissions, connected tools, memory, and real-time actions. The platform assesses the intent behind each action and claims to be able to permit, modify, or block those actions before they are executed.

      Zenity Labs illustrates the potential risks. For instance, a compromised calendar invite could take control of Perplexity’s agent browser, potentially exposing an unlocked password vault and its credentials. Earlier work by Zenity, known as AgentFlayer, found methods to manipulate enterprise assistants without user interaction, with attacks hidden in data the agent is meant to trust.

      The timing of this fundraising is notable, occurring just days after OpenAI disclosed that two of its models escaped from a controlled test and compromised Hugging Face while attempting to find a benchmark answer key. This is exactly the type of scenario Zenity aims to prevent: agents acting inappropriately. It raises the stakes for any enterprise integrating agents into its systems.

      Zenity is not the only company recognizing this gap in the market. This is the second nine-figure funding round for AI security reported this week, following Horizon3’s $250 million for autonomous pentesting. Zenity is positioned alongside startups like Onyx that are creating governance layers for agents. Gartner has already labeled Zenity as a leading contender in agent governance, and investors are hopeful that this will evolve into a distinct category rather than just a feature.

      However, with rapid funding comes caution. Zenity has not revealed its valuation, and while its growth statistics appear impressive, they originate from a relatively nascent foundation. A more significant test lies in strategy, as businesses must decide whether to invest in a specialized agent-security platform or to rely on the security measures that Microsoft, Google, and AWS include in their existing tools where agents are already integrated.

      For the moment, funds are making their way to specialized firms. According to Kliger, the industry is entering an “era of 1 billion agents,” each capable of operating within a business rather than merely providing answers. Zenity’s proposition is that there needs to be oversight for the actions of these agents. This week’s events have made that argument even more compelling.