Prisma AIRS vs Zenity

Prisma AIRS is Palo Alto Networks' platform for securing AI applications, agents, models and data. Palo Alto Networks launched it on April 28, 2025, the same day it announced its intent to acquire Protect AI, a Seattle AI security company, and completed that acquisition on July 22, 2025. Prisma AIRS 2.0, released October 28, 2025, "completes the native integration" of Protect AI, and Protect AI's chief executive, Ian Swanson, joined as VP, Product, Prisma AIRS. The protectai.com site now redirects to the Prisma AIRS page, and the README of Protect AI's open-source LLM Guard library says the project has been archived. Palo Alto Networks was incorporated in 2005 and is based in Santa Clara; this profile lists the company's founding year and headquarters, as the directory does for other big-company products. Palo Alto's fiscal 2026 annual report lists six parts: AI Gateway, Agent Security, AI Red Teaming, AI Runtime Security, AI Model Security (labeled AI Supply Chain Security on the current product page) and AI Posture Management. AI Runtime Security comes in two forms. API Intercept is called from application code through a REST API to scan prompts and model responses. Network Intercept inspects cloud network traffic inline and is managed from Strata Cloud Manager or Panorama. The platform supports AWS, Azure, Google Cloud, Kubernetes, private clouds and on-premises infrastructure. AI Red Teaming tests agentic and non-agentic systems with a library of more than 500 attack vectors across 50+ techniques and produces a risk score. AI Model Security scans models for tampering, malicious code, data poisoning, architectural backdoors and unsafe serialization formats. **Agent security** The product page now adds Agent Identity Security and Agentic Endpoint Security. Palo Alto completed its acquisition of Koi, which adds agentic endpoint security, in April 2026. On May 29, 2026 it completed its acquisition of Portkey, whose gateway becomes the core AI Gateway for Prisma AIRS: it routes, monitors and secures traffic between agents and models. **Frameworks** Palo Alto maps AI Red Teaming attacks to the OWASP Top 10 for LLMs, MITRE ATLAS and what it calls the NIST Risk Management Framework. We found no documented support for the EU AI Act or ISO 42001. **Who it's for** Enterprises, especially existing Palo Alto Networks customers on Strata Cloud Manager, that want model scanning, red teaming and runtime protection funded from the same credit pool as their software firewalls.

Zenity is a security and governance platform for AI agents: agents inside SaaS tools, agents built in-house on cloud AI platforms, and coding and personal agents on employee devices. It was co-founded by Ben Kliger (CEO) and Michael Bargury (CTO). It launched in April 2021 as a security platform for low-code/no-code development, added support for enterprise AI copilots in November 2023 and now focuses on agents. Research and development is in Tel Aviv, and go-to-market and operations are led from New York. Zenity says it has more than 230 employees. Its company page lists a $5 million seed round (November 2021), a $16.5 million Series A led by Intel Capital (September 2023), a strategic investment from Microsoft's M12 (July 2024), a $38 million Series B co-led by Third Point Ventures and DTCP (October 2024) and a $125 million Series C led by Norwest (August 2026). Zenity released a new version of its platform in August 2026, organized in three parts. Surface covers AI Observability (an inventory of agents, MCP servers, tools and data access, including agents built without security review), AISPM (configuration and permission checks before agents go live) and Exposure Management (which access paths are actually exploitable). Enforce covers Runtime Boundaries, which checks each agent action in real time and allows it, blocks it or shuts the agent down, including multi-step chains such as data exfiltration or privilege escalation; Agentic Identity, which gates actions on the identity an agent acts for in Okta or Microsoft Entra ID; and MCP Security, a gateway that allows, modifies or blocks tool calls and flags tool poisoning, rug pulls, shadow servers and over-scoped servers. Protect is AIDR, which detects direct and indirect prompt injection, data exfiltration, tool misuse and memory poisoning with LLM-based intent detection. Zenity names Microsoft 365 Copilot, Copilot Studio, Power Platform, ChatGPT Enterprise, Claude Enterprise, Salesforce Agentforce, ServiceNow, Amazon Bedrock and AgentCore, Google Vertex AI and Microsoft Foundry among the platforms it covers. For coding agents such as Cursor, Claude Code, Codex and GitHub Copilot, it collects activity through native agent hooks and OpenTelemetry and enforces policy through its MCP gateway. It sends agent findings to AWS Security Hub (in OCSF format) and to ServiceNow SecOps. **AI compliance coverage** Zenity maps its detections to the OWASP LLM Top 10 and MITRE ATLAS. We found no mapping on its site to the EU AI Act, the NIST AI RMF or ISO 42001. Its trust center says Zenity holds SOC 2 Type II and is ISO 27001 compliant and ISO 27701 compliant; reports are available on request. **Who it's for** Large enterprises running many agents across Microsoft, Salesforce, ServiceNow and OpenAI platforms, plus homegrown and coding agents, that want one policy layer over all of them. Zenity says most of its customers are Fortune 500 or Global 2000 companies; SoftBank Corp. is the only one named, in its Series C announcement. Gartner named Zenity a Cool Vendor in Agentic AI TRiSM (September 2025), and Zenity quotes an April 2026 Gartner report calling it the company to beat in AI agent governance.