Zenity is an Agentic AI Security vendor based in New York, USA and Tel Aviv, Israel. Its flagship product is Zenity AI Security Platform (AI Observability, AISPM, Exposure Management, Boundaries, Agentic Identity, MCP Security, AIDR). Against the OWASP LLM Top 10 (2026) we score it Covered on 4 of 10 risks, Partial on 4, with 6 MITRE ATLAS techniques mapped as of . It's independent.
- 4/10
- Risks covered
- 4
- Partial
- 6
- ATLAS techniques
- $180M+
- Funding
- Active
- Status
At a glance
- Category
- Agentic AI Security
- Headquarters
- New York, USA and Tel Aviv, Israel
- Founded
- 2021
- Employees
- 201-500
- Flagship product
- Zenity AI Security Platform (AI Observability, AISPM, Exposure Management, Boundaries, Agentic Identity, MCP Security, AIDR)
- Deployment
- Posture scanning and runtime detection across SaaS, cloud and endpoint agents; inline prevention on Copilot Studio, Microsoft Foundry and coding agents via hooks; detection and posture only on poll-only platforms such as M365 Copilot, Agentforce, Vertex AI, Gemini Enterprise and ServiceNow
- Best for
- Enterprises running agents across Copilot, ChatGPT Enterprise, Agentforce and coding tools that need one policy layer over agent actions
- Funding to date
- $180M+
- Last round
- Series C ($125M) · 2026-08
- Status
- Active
Facts as of .
What does Zenity do?
Governs what enterprise agents do, not what they say: posture scanning plus runtime blocking of the action before it executes. It ships as Posture scanning and runtime detection across SaaS, cloud and endpoint agents; inline prevention on Copilot Studio, Microsoft Foundry and coding agents via hooks; detection and posture only on poll-only platforms such as M365 Copilot, Agentforce, Vertex AI, Gemini Enterprise and ServiceNow, and it's built for enterprises running agents across Copilot, ChatGPT Enterprise, Agentforce and coding tools that need one policy layer over agent actions. It's a weaker fit for teams that need rate limits or cost guardrails, hallucination checks, or training-pipeline and model-artifact monitoring; Zenity states none of these is a Zenity control.
Which OWASP LLM Top 10 risks does Zenity cover?
Zenity's strongest verdict is LLM01 Prompt Injection. The two tables below come from our editorial read of Zenity's public documentation as of (it's what the docs support, not a hands-on lab test), and this row also sits in the full matrix across all 27 vendors.
| OWASP LLM Risk | What it is | Zenity coverage | How it's addressed | Source |
|---|---|---|---|---|
| LLM01 Prompt Injection | User or hidden input overrides the model's rules or intended behavior. | Covered | AIDR detects direct and indirect injection, and Inline Prevention blocks the resulting action on Copilot Studio, Microsoft Foundry and coding agents. | https://zenity.io/blog/how-zenity-implements-the-owasp-top-10-for-llm-applications |
| LLM02 Sensitive Information Disclosure | The model leaks secrets, personal data, or proprietary content in its output. | Covered | AIDR detects PII, financial data and secrets, and Inline Prevention Sensitive Data Blocking stops inference-time leaks before delivery. | https://zenity.io/blog/how-zenity-implements-the-owasp-top-10-for-llm-applications |
| LLM03 Excessive Agency | An agent holds more permissions, tools, or autonomy than the task needs. | Covered | AISPM surfaces over-provisioned agent access, and Inline Prevention Tool Misuse Prevention blocks unsafe tool calls at runtime. | https://zenity.io/blog/how-zenity-implements-the-owasp-top-10-for-llm-applications |
| LLM04 Supply Chain | Compromised models, datasets, plugins, or dependencies add risk before runtime. | Partial | AISPM inventories and risk-scores third-party AI integrations and discovers MCP servers, which Zenity labels posture only. | https://zenity.io/blog/how-zenity-implements-the-owasp-top-10-for-llm-applications |
| LLM05 Data and Model Poisoning | Tampered training or fine-tuning data corrupts how the model behaves. | Partial | AIDR flags malicious input at runtime, and Zenity states it does not monitor training or model-artifact pipelines. | https://zenity.io/blog/how-zenity-implements-the-owasp-top-10-for-llm-applications |
| LLM06 Unbounded Consumption | Uncontrolled requests drive cost, denial of service, or model extraction. | Not covered | Zenity states it has no AI-specific rate limit, inference ceiling or cost guardrail. | https://zenity.io/blog/how-zenity-implements-the-owasp-top-10-for-llm-applications |
| LLM07 Misinformation | The model produces false or fabricated content that users act on. | Not covered | Zenity states hallucination and factual accuracy are not a Zenity control, and Risk Filtering covers toxicity and unsafe content only. | https://zenity.io/blog/how-zenity-implements-the-owasp-top-10-for-llm-applications |
| LLM08 Hidden Context Exposure | Hidden instructions, retrieved policy text, or tool schemas get extracted, inferred, or reconstructed from the model. | Partial | AIDR output monitoring and secrets detection catch some context exposure, and Zenity says full system-prompt-extraction defense is broader than the product. | https://zenity.io/blog/how-zenity-implements-the-owasp-top-10-for-llm-applications |
| LLM09 Vector and Embedding Weaknesses | Flaws in RAG stores let attackers poison, extract, or infer data. | Partial | Zenity detects anomalous data access at runtime, but says vector-store and embedding integrity is not a core control. | https://zenity.io/blog/how-zenity-implements-the-owasp-top-10-for-llm-applications |
| LLM10 Improper Output Handling | Downstream systems trust model output without checking it, enabling injection or code execution. | Covered | AIDR detects malicious links, hidden text and risky image rendering, and Inline Prevention evaluates output before it reaches downstream tools. | https://zenity.io/blog/how-zenity-implements-the-owasp-top-10-for-llm-applications |
| ATLAS Technique (ID) | Tactic | Zenity coverage | Notes | Source |
|---|---|---|---|---|
| LLM Prompt Injection (AML.T0051) | Execution | Covered | AIDR detects direct and indirect prompt injection and blocks execution. | https://zenity.io/platform/ai-detection-and-response |
| Exfiltration via AI User Interface (AML.T0136) | Exfiltration | Covered | AIDR monitors and blocks sensitive data leaving through agent conversations, tool calls or encoded payloads. | https://zenity.io/platform/ai-detection-and-response |
| Data from AI Services: System Instructions (AML.T0085.003) | Collection | Partial | AIDR output monitoring catches some exposure, and full system-prompt-extraction defense is outside the product. | https://zenity.io/blog/how-zenity-implements-the-owasp-top-10-for-llm-applications |
| AI Agent Tool Invocation (AML.T0053) | Execution / Privilege Escalation / Lateral Movement | Covered | Boundaries evaluates every agent action in real time and can block it or shut the agent down. | https://zenity.io/platform/runtime-boundaries |
| AI Agent Tool Poisoning (AML.T0110) | Persistence | Covered | MCP Security names tool poisoning and rug pulls among governed risks and can allow, modify or block a call on live context. | https://zenity.io/platform/mcp-security |
| Supply Chain Compromise (AML.T0010) | Initial Access | Partial | AISPM inventories and risk-scores third-party integrations, which Zenity labels posture only. | https://zenity.io/blog/how-zenity-implements-the-owasp-top-10-for-llm-applications |
Which OWASP Agentic risks does Zenity cover?
5 of 10 Covered and 4 Partial against the OWASP Top 10 for Agentic Applications, scored from public documentation as of 2026-10-30.
| Agentic Risk | What it is | Zenity coverage | How it's addressed | Source |
|---|---|---|---|---|
| Agent Goal Hijack (ASI01) | Attackers redirect what an agent is trying to do, often through documents or other content it reads. | Covered | AIDR detects direct and indirect prompt injection and blocks execution before the agent acts on it. | https://zenity.io/platform/ai-detection-and-response |
| Tool Misuse and Exploitation (ASI02) | An agent uses a legitimate tool in a harmful way, or a manipulated tool call does damage. | Covered | AIDR flags and blocks unauthorized tool invocations or API calls that violate policy. | https://zenity.io/platform/ai-detection-and-response |
| Identity and Privilege Abuse (ASI03) | Agents inherit, reuse, or escalate credentials and permissions beyond what the task needs. | Covered | Agentic Identity pulls identities from Entra ID and Okta and gates actions at runtime, such as blocking a deactivated Okta user. | https://zenity.io/platform/agentic-identity-and-access-management |
| Agentic Supply Chain Vulnerabilities (ASI04) | Compromised MCP servers, plugins, skills, or agents enter the runtime. | Covered | Zenity runs risky skills, files and code in a sandbox before they reach the environment, and MCP Security blocks risky calls at its gateway. | https://zenity.io/use-cases/agent-type/coding-personal-agents |
| Unexpected Code Execution (ASI05) | An agent generates or runs code that gives an attacker execution. | Partial | Native hooks and the MCP gateway block or modify a dangerous tool call before it executes, but no code-level inspection of generated code is documented. | https://zenity.io/use-cases/agent-type/coding-personal-agents |
| Memory and Context Poisoning (ASI06) | Attackers plant content in an agent's memory or context that shapes later decisions. | Partial | AIDR detects manipulation embedded in retrieved content or tool results, and blocking of memory writes is not documented. | https://zenity.io/platform/ai-detection-and-response |
| Insecure Inter-Agent Communication (ASI07) | Messages between agents get spoofed, intercepted, or tampered with. | Partial | AIDR gives agent-to-agent visibility into requests and responses, but no message authentication or anti-spoofing control is documented. | https://zenity.io/platform/ai-detection-and-response |
| Cascading Failures (ASI08) | One fault spreads across connected agents and systems. | Partial | Boundaries stops multi-step exfiltration and privilege-escalation chains, but containment across connected agents is not documented. | https://zenity.io/platform/runtime-boundaries |
| Human-Agent Trust Exploitation (ASI09) | An agent's confident output talks a person into approving something harmful. | Unverified | AISPM can force end-user confirmation on sensitive Copilot Studio actions, but no control against manipulated human approval is named. | https://zenity.io/blog/how-zenity-implements-the-owasp-top-10-for-llm-applications |
| Rogue Agents (ASI10) | An agent drifts or is compromised and acts outside policy. | Covered | Boundaries evaluates every agent action in real time and lets it through, blocks it, or shuts the agent down. | https://zenity.io/platform/runtime-boundaries |
Is Zenity independent, and how is it funded?
Zenity is an independent company as of . It has raised $180M+ to date, most recently a Series C ($125M) dated 2026-08. Lead investors: Norwest.
Zenity alternatives
The closest alternatives we track in Agentic AI Security are AIR Security, Akto, Knostic. On the open-source side, agent-scan covers similar ground.
Frequently asked questions
What is Zenity used for?
Zenity is an Agentic AI Security vendor. Its flagship product is Zenity AI Security Platform (AI Observability, AISPM, Exposure Management, Boundaries, Agentic Identity, MCP Security, AIDR). It ships as Posture scanning and runtime detection across SaaS, cloud and endpoint agents; inline prevention on Copilot Studio, Microsoft Foundry and coding agents via hooks; detection and posture only on poll-only platforms such as M365 Copilot, Agentforce, Vertex AI, Gemini Enterprise and ServiceNow, and it's built for enterprises running agents across Copilot, ChatGPT Enterprise, Agentforce and coding tools that need one policy layer over agent actions.
Is Zenity independent or acquired?
Zenity is an independent company as of 2026-10-30 and has not been acquired.
How many OWASP LLM Top 10 risks does Zenity cover?
We score Zenity as Covered on 4 of the ten OWASP LLM Top 10 (2026) risks, with 4 more Partial, as of 2026-10-30. The table on this page breaks down every risk, including the ones marked Not covered or Unverified.
How is Zenity deployed?
Zenity ships as Posture scanning and runtime detection across SaaS, cloud and endpoint agents; inline prevention on Copilot Studio, Microsoft Foundry and coding agents via hooks; detection and posture only on poll-only platforms such as M365 Copilot, Agentforce, Vertex AI, Gemini Enterprise and ServiceNow.
What are the best alternatives to Zenity?
The closest Agentic AI Security alternatives llmthreat tracks are AIR Security, Akto, Knostic. On the open-source side, agent-scan covers similar ground.