The race to govern AI agents just got a serious injection of capital. Zenity, a Tel Aviv and New York-based startup that builds security and governance tooling for enterprise AI agents, closed a $125 million Series C in early August 2026, led by Norwest Venture Partners.

New investors Qumra Capital, SoftBank Vision Fund 2, Hitachi Ventures, and LG Technology Ventures joined the round. Existing backers Vertex Ventures, Third Point Ventures, DTCP, and Intel Capital also participated. The raise brings Zenity’s total funding to roughly $185 million and accelerates a pattern that’s been building all year: investors are pouring money into the layer that watches what AI agents actually do.

The Problem Zenity Is Solving

Most conversations about enterprise AI focus on what agents can do. Zenity is focused on what they might do that you didn’t intend.

AI agents don’t just answer questions anymore. They access file systems, call APIs, write code, send emails, and make decisions on behalf of employees across multi-step workflows. When agents operate at that level of autonomy, the traditional security model, which was built for humans accessing systems, breaks down fast.

Founded in 2021 by Unit 8200 veterans Ben Kliger (CEO) and Michael Bargury (CTO), who previously worked together on cloud and operational technology security at Microsoft, Zenity built its platform specifically for this new attack surface. Its approach spans three tiers: SaaS-managed agents like Microsoft 365 Copilot and Salesforce Agentforce; cloud-built agents running on AWS Bedrock, Azure AI Foundry, and Google Vertex AI; and endpoint agents like GitHub Copilot, Cursor, and Claude Desktop.

The platform discovers all agents operating in a business environment, including shadow agents that departments have built without IT sign-off. It maps what data each agent touches, what tools it can invoke, and what permissions it holds. Then it enforces policy in real time.

Runtime Boundaries: Governing Before the Damage Happens

In July 2026, Zenity introduced what it calls Runtime Boundaries, a new security architecture that steps in before an agent action executes rather than after. Traditional security posture scans are periodic snapshots. Zenity’s approach is event-driven, reflecting configuration, permission, and connector changes within minutes of occurrence.

This matters because the failure mode for AI agents isn’t usually a dramatic hack. It’s an agent that had permission to read HR data for one workflow gradually accumulating access across more systems as it tries to complete tasks. By the time someone notices, sensitive information has already moved.

Runtime Boundaries let security teams define what agents can and cannot do at a granular level, drawing from risk frameworks like OWASP LLM Top 10 and MITRE ATLAS. Enforcement happens before the action takes effect, not after.

In April 2026, Gartner named Zenity the “Company to Beat” in AI agent governance, a designation that reflects both how seriously the analyst firm takes the space and how far ahead Zenity has moved compared to competitors.

Why This Matters for Businesses Deploying AI Agents

If you’re running AI agents in your business, whether through Microsoft Copilot, a custom-built solution on Bedrock or Vertex, or a third-party agentic platform, you now have a sprawling agent estate that most IT and security teams cannot see clearly.

The funding signals what the market already knows: governance is not optional as agents get more capable. Organizations that treat AI security as an afterthought will eventually discover an agent did something unexpected with production data or a customer system.

The specific risks worth understanding:

Privilege escalation. An agent with access to a CRM might gradually accumulate access to financial systems if its policy boundaries are loosely defined. Agents are goal-seeking, and they’ll use whatever access they can find.

Shadow agents. Business units stand up AI tools faster than IT can track. Without a discovery layer, the security team has no visibility into agents that marketing, finance, or operations built on a no-code platform.

Prompt injection. Malicious instructions embedded in data that agents process, a poisoned email, a compromised document, can redirect agent behavior in ways the original developer never anticipated.

Excessive data access. Agents often request broad permissions during setup that were never intended for production use. Without governance tooling, those permissions persist indefinitely.

What This Means for Business

The $125M raise tells you something important: the enterprises writing checks are not waiting for a mature governance market to develop. They’re investing in governance now, alongside their agent deployment, not after the fact.

For business leaders, the practical takeaway is straightforward. Every AI agent you deploy is, in effect, a digital employee with system access. You’d never hire an employee and give them unrestricted access to all your business systems with no oversight. AI agents deserve the same thoughtful access controls.

The good news is that the tooling is maturing fast. Platforms like Zenity now offer the same governance visibility for AI agents that companies have long had for human users. The cost of getting governance right early is far lower than cleaning up an agent-related incident after the fact.

At Enterprise DNA, we work with businesses at every stage of their AI agent journey. Governance architecture is one of the first conversations we have in any Omni advisory engagement, because agents that aren’t governed properly undermine the confidence that makes the rest of the deployment worthwhile. If you’re thinking about deploying AI agents at scale, get in touch with our team to talk through how governance fits into the picture.