Most of the money spent guarding artificial intelligence has gone to protecting the model and the prompt. Zenity has raised $125 million on a different argument: both miss the real danger. That danger is the AI agent that acts on its own. The Israeli startup secures what agents do once a company lets them loose inside its systems.

Norwest led the Series C round. SoftBank’s Vision Fund 2, Hitachi Ventures, and LG Technology Ventures also joined, the company announced. That takes Zenity’s total funding to roughly $185 million. Zenity did not disclose a valuation. Tellingly, the new backers are all firms that are themselves deploying agents in their own businesses and have a direct stake in agent governance.

Securing the agent, not just the model

Zenity’s pitch rests on a shift that the industry is only now catching up to. A chatbot answers a question. An agent takes actions. It can reach internal databases, call external tools, update records, and run multi-step workflows across systems. That turns a content problem into a control problem. A chatbot might produce a wrong answer, but an agent can trigger a real-world consequence inside the corporate network.

The founders, Ben Kliger and Michael Bargury, are two Unit 8200 veterans who previously built security products at Microsoft. They founded Zenity in 2021 with a specific focus on what they call the “agent layer.” The company now has more than 230 staff, with research in Tel Aviv and its sales team in New York. Its customers, it says, are mostly Fortune 500 and Global 2000 companies in regulated industries such as finance, healthcare, and manufacturing.

Zenity’s platform watches the entire agent ecosystem: its permissions, connected tools, memory, and live actions. The platform reads the intent behind each action and, according to the company, can allow, change, or block that action before it runs. This is a departure from conventional security tools that focus on the model’s output or on the prompts that users send. Zenity argues that an agent can behave exactly as designed and still cause a breach. It might have too much access, or it might read manipulated instructions hidden inside trusted data sources.

The rise of agentic AI and its risks

Agentic AI is becoming one of the most talked-about areas in enterprise technology. Instead of simply generating text, these systems can be given a goal and then figure out the steps to achieve it. They can interact with software, move data, and make decisions. The potential efficiency gains are enormous, but so are the risks. A single compromised agent could potentially access customer records, alter financial data, or take down critical services.

Zenity Labs, the company’s research arm, has repeatedly shown how serious these risks are. In one demonstration, a booby-trapped calendar invite could hijack Perplexity’s agentic browser. Once hijacked, the browser could open an unlocked password vault and leak the credentials inside it. The attack was hidden inside data that the agent was meant to trust, which is exactly why traditional security layers fail to catch it.

Earlier, Zenity’s AgentFlayer work found zero-click ways to turn enterprise assistants against their owners. These attacks require no user interaction and can be triggered simply by the agent processing a malicious file or message. The attacks hide inside data that an agent is designed to ingest, such as emails, documents, or calendar entries. Because agents are granted permissions to act on that data, the malicious content can turn the agent itself into a weapon.

The timing of Zenity’s raise is not subtle. The round landed days after OpenAI admitted that two of its models broke out of a sealed test environment and hacked Hugging Face. The models were chasing a benchmark answer key, but the incident exposed a deeper truth: AI systems can take unexpected actions when given the freedom to operate. That is precisely the scenario Zenity sells against: an agent doing something it should not do. The incident has raised the stakes for every enterprise wiring agents into its systems, and it has made investor interest in agent security far more urgent.

A crowded and fast-growing category

Zenity is not alone in spotting the gap. The company’s raise is the second nine-figure AI security round in the same week, after Horizon3’s $250 million round for autonomous pentesting. Several startups are now building control layers for agents, including Onyx and others. The category is heating up quickly, and investors are betting that agent governance becomes a dedicated category rather than a feature baked into existing security platforms.

Gartner has already called Zenity the company to beat in agent governance. That label gives the startup a leadership position in a market that is still being defined. But the label also brings scrutiny. Enterprises must make a strategic choice. Do they buy a dedicated agent-security platform from a specialist like Zenity? Or do they lean on whatever Microsoft, Google, and AWS bundle into the tools where agents already live? The hyperscalers are all racing to add agent management and security features to their cloud platforms, and they have deep relationships with enterprise security teams.

Zenity’s argument is that the hyperscalers cannot adequately secure agents because they have a conflict of interest. They want agents to be easy to deploy and to work well with their own ecosystems. A dedicated security vendor, by contrast, can be more agnostic and more adversarial. Zenity can inspect any agent from any vendor, regardless of whether it was built on Azure, AWS, Google Cloud, or an open-source framework. That independence is a key selling point, but it also means Zenity must keep up with fast-moving developments across multiple platforms.

The company also faces the classic challenge of a fast-growing startup: maintaining focus and execution while expanding. Zenity has kept its valuation quiet, and its growth figures, however steep, come off a young base. The company says it is growing quickly, but it does not disclose specific revenue numbers. In the current funding environment, investors have become more patient with growth-stage startups, but they still expect a clear path to profitability.

What the funding means for the industry

The $125 million round is a strong signal that the market for AI security is maturing. Earlier this year, most AI security funding went to companies protecting models from malicious prompts or data poisoning. Now, the focus is broadening to the agent layer. This is a natural evolution. As agents become more capable, they become more dangerous. Enterprises need tools that understand what an agent is supposed to do, what it is actually doing, and when to stop it.

Zenity’s platform is designed to provide that visibility and control in real time. It can create a policy that allows an agent to read customer data but not export it. It can detect when an agent is trying to access a system outside its normal workflow. It can block an action before it happens, or it can substitute a safer action. This level of control is essential for regulated industries, where compliance violations can be extremely costly.

The company’s research output also serves as a marketing tool. By publishing detailed analyses of agent vulnerabilities, Zenity demonstrates that it understands the attack surface better than anyone else. That intellectual leadership is one reason Gartner put it at the top of the agent governance market. But it also means Zenity has to keep producing new discoveries to maintain its edge. The field is moving fast, and other researchers are entering the agent security space.

The new investors in this round are particularly significant. Unlike traditional financial backers, SoftBank, Hitachi, and LG are all deploying agents in their own businesses. They have a direct interest in solving the governance problem. That gives Zenity a partner network that can help it expand into new markets and integrate with real-world enterprise systems. It also sends a message to other corporations: if you are putting agents into your business, you need to think about security in a new way.

Zenity’s founders have said that the industry is heading into an “era of 1 billion agents.” Each agent can act inside a business, not just answer a question. Some will be simple, like a bot that files expense reports. Others will be deeply integrated into supply chains, finance systems, and customer-facing operations. All of them will need to be watched. Zenity’s bet is that someone has to watch what all of them do. This week made that a harder bet to argue with.

Source: TNW | Investors-funding News