Zenity

Primary category: agent-runtime-security. Also spans ai-spm.

One-liner — An end-to-end security and governance platform for AI agents, enterprise copilots, and low-code/no-code apps — giving security teams visibility into and control over what agents can do.

What it does

Zenity discovers and inventories the sprawl of AI agents, copilots, and low-code/no-code automations across an enterprise (it cites large enterprises averaging ~80,000 of them), assesses their risk and misconfigurations, and governs/protects their behavior at runtime. It made its name finding that a large share of copilots and low-code apps ship with security vulnerabilities, and on offensive research showing how agents can be hijacked (prompt injection, “living off AI,” AgentFlayer). The product combines AI-SPM (discovery + posture for agents/copilots) with agent runtime protection and governance.

Where it sits in the stack

Primary agent-runtime-security (runtime visibility + control over agent behavior), with deep overlap into ai-spm (inventory + posture of the agent/copilot estate). Layer: model/prompt. It addresses all three lethal-trifecta legs for agents — detects untrusted input (prompt injection into copilots/agents), watches sensitive data agents can reach, and constrains agent actions/egress. The Zenity/Noma pair is the canonical example of the AI-SPM ↔ agent-runtime overlap in this taxonomy.

Deployment & architecture

SaaS control plane with API integrations into the agent/copilot platforms it secures — Microsoft Copilot Studio / Power Platform, Salesforce, ServiceNow, OpenAI/Anthropic-based agents, and other low-code/agent ecosystems — plus connectors for posture scanning and runtime monitoring of agent activity. Findings export to SIEM/SOC. Strong Microsoft alignment (M12 is an investor; CEO is ex-Microsoft).

Positioning & differentiators

Known as an early, research-forward leader in AI agent and copilot security — CTO Michael Bargury’s offensive research (Black Hat / DEF CON talks on copilot and agent exploitation) is core to the brand. Differs from noma-security (which came up via ML/data-pipeline AI-SPM) by its origin in low-code/copilot security and its runtime-governance emphasis. Differs from runtime AI-firewall players like prisma-airs, straiker, and operant-ai by focusing on the agent/copilot control plane (what agents are configured to do and access) rather than just inline prompt/response filtering.

Ownership, funding & M&A

Independent, VC-backed. Founded 2021 by Ben Kliger (CEO, ex-Microsoft cloud workload protection product leader) and Michael Bargury (CTO). $38M Series B (Oct 29, 2024) co-led by Third Point Ventures and DTCP, with Microsoft’s M12, Intel Capital, and Vertex Ventures; total raised over $55M. No M&A; no seed acquisition flag. Ownership confidence: high.

CTO / hedge-fund lens

Day-2, becoming Day-1 fast for shops deploying Microsoft Copilot / Copilot Studio, Power Platform, or other low-code agents at scale — exactly the “citizen developer builds an agent that can touch sensitive data” risk. For a hedge fund running M365 Copilot or building internal agents, Zenity is one of the more relevant agent-security names; it gives the SOC an inventory and guardrails over agent behavior, useful for both security and SR 11-7-style model/agent oversight. Less relevant if you have essentially no agent/copilot footprint.

Competitors / alternatives

noma-security, straiker, operant-ai, lasso-security, prisma-airs, reco, apex-security.

Open questions / to verify

• Runtime enforcement depth — does it block agent actions inline, or mostly detect + govern via the platform APIs?
• Coverage beyond the Microsoft/low-code ecosystem for code-built agents.
• Any funding since the Oct 2024 Series B.

Sources

• Zenity Raises $38M Series B (zenity.io newsroom) — fetched 2026-06-28 — supports: founders, HQ, Series B $38M/date/investors, total funding, product scope, customers; confidence: high (vendor release).

History

• [2026-06-28] Stub created from seed registry.
• [2026-06-28] Researched; established independent/VC-backed, founded 2021 (Kliger/Bargury, Tel Aviv), $38M Series B (2024-10, Third Point & DTCP), $55M+ total. Confirmed agent-runtime + AI-SPM dual scope (the canonical AI-SPM/agent-runtime overlap with Noma). Set ownership_confidence high.