Zenity

Zenity began in 2021 as a security and governance company for low-code and no-code development, then expanded that same control problem into enterprise AI agents. The product is now presented as an agent-centric security platform with three connected functions: surface the agents and applications an organization has, enforce guardrails over what they can access and do, and protect them through detection and response. This is more specific than generic AI governance. Zenity focuses on the execution layer, including agent identity, tools, permissions, memory, data access, control flow, and the sequence of actions that leads to an outcome. Its current coverage claims include Microsoft Copilot and Copilot Studio, Salesforce Agentforce, AWS Bedrock, Google Vertex AI, ChatGPT Enterprise, Claude Enterprise, ServiceNow, Power Platform, and coding assistants such as GitHub Copilot, Cursor, and Claude Code.

The customer problem is credible and expanding. Enterprises are deploying agents made by professional developers, business users, and employees using low-code platforms, often across environments that were previously governed by separate AppSec, identity, endpoint, data-loss-prevention, and cloud tools. An inventory of models or prompts is insufficient when an agent can retain context, call external tools, read sensitive records, execute code, or chain actions over time. Zenity's proposed answer is continuous visibility and policy at the decision point: identify who owns an agent, what it can reach, how it behaves, whether the action is consistent with its intended task, and whether to allow, flag, redact, or block it. The low-code heritage is strategically useful because citizen-built automations remain a major source of undocumented applications and over-broad access, but it also creates a positioning challenge as the company competes for budgets labeled AI security, application security, identity, or data security.

Zenity has stronger commercialization signals than the prior Series A label implied. Its official timeline records a $5M seed in 2021, a $16.5M Series A led by Intel Capital in September 2023, a strategic investment from Microsoft's M12 in July 2024, and a $38M Series B co-led by Third Point Ventures and DTCP in October 2024. The company says it serves large enterprises across financial services, technology, manufacturing, energy, and pharmaceuticals, and its website publishes anonymized customer outcome metrics; these are company-reported signals rather than independently audited revenue or retention evidence. A current LinkedIn company profile places Zenity in the 201–500 employee band and identifies New York as its primary headquarters with a Tel Aviv location. Current hiring and product material suggest an international go-to-market operation, but public sources do not establish exact headcount, ARR, customer concentration, renewal rates, or deployment scale.

The competitive field is converging quickly. Zenity competes with AI security and AI-SPM vendors such as Protect AI, HiddenLayer, Lakera, Prompt Security, and Noma Security, while also facing substitutes from Microsoft, cloud providers, identity vendors, EDR/XDR platforms, DSPM products, and internal security engineering. Its potential edge is the combination of agent inventory, low-code governance, execution-path telemetry, and preventative runtime controls across SaaS, cloud, and endpoint surfaces. Its AgentFlayer research and public work on prompt injection and agent behavior may strengthen credibility with security practitioners, but research visibility is not the same as durable product differentiation. The key diligence question is whether customers receive reliable cross-platform enforcement and measurable reduction in risky actions, or mainly another discovery and alerting layer.

For defense, intelligence, and critical-infrastructure users, the dual-use case is substantive but indirect. The platform could help govern copilots, coding agents, workflow automation, and mission-support applications that handle sensitive documents, credentials, operational data, or privileged tools. Context-aware monitoring and the ability to interrupt unsafe actions are relevant where an accidental disclosure, poisoned context, or destructive autonomous step can have operational consequences. However, public evidence does not show a defense contract, classified deployment, security accreditation, or mission-specific integration. Strategic relevance therefore rests on the transferability of commercial agent controls into high-assurance environments, subject to procurement, isolation, logging, supply-chain, and authorization requirements that are not demonstrated in the public record.