Six vendors · Alphabetical
GenAI security vendor profiles: what six vendors state about AI controls
In short
Six vendors that sell network security from the cloud and publish AI security capabilities, profiled from their own pages. Each profile lists what the vendor describes for the three approaches and thirteen checks, its stated responses, its strengths and at least two limits. They are in alphabetical order; the quiz orders them by fit to your answers.
Cato Networks
Cato AI Security on the Cato SASE Cloud
Cato Networks states Watch, Redact and Block; Coach is not in our sources.
States 10 of 13 checks, partly 1
Strength: One engine and one console for employee AI use, homegrown AI apps and AI agents, on the platform that also carries network and security policy, as Cato describes it.
Limit: FedRAMP is not listed on Cato's compliance page, which matters for US federal buyers.
Read the profileCisco
AI Access in Cisco Secure Access, and Cisco AI Defense
Cisco states Block; Watch and Coach are partly stated; Redact is not in our sources.
States 4 of 13 checks, partly 5
Strength: AI Access is part of Cisco Secure Access, managed from one console with centralized policy.
Limit: Prompt-level DLP, named coding agents and MCP are not described on the pages we read.
Read the profileFortinet
FortiAI and FortiAIGate; FortiSASE
Fortinet states Watch, Coach and Block; Redact is not in our sources.
States 8 of 13 checks, partly 2
Strength: Self-hosted AI enforcement: with FortiAIGate self-hosted, data and logs stay on your own infrastructure.
Limit: Which enforcement point delivers employee AI controls (FortiSASE, a FortiGate or FortiAIGate) is not stated on the pages we read.
Read the profileNetskope
Netskope One and the Netskope AI products (AI Command Center, GenAI App Security, Agentic Broker, AI Gateway, AI Guardrails, Agent Action Control, AI Red Teaming)
Netskope states Block; Coach is partly stated; Watch and Redact are not in our sources.
States 5 of 13 checks, partly 5
Strength: A dedicated Agentic Broker that decodes MCP traffic and blocks unauthorized MCP communication.
Limit: General GenAI app discovery and a watch-only action are not detailed on the pages we read; the stated discovery centres on MCP.
Read the profilePalo Alto Networks
AI Access Security (Prisma SASE) and Prisma AIRS
Palo Alto Networks states Coach and Block; Watch and Redact are not in our sources.
States 7 of 13 checks, partly 3
Strength: Two named products for two jobs: AI Access Security in Prisma SASE for employee GenAI use, and Prisma AIRS for agents, apps and models.
Limit: MCP traffic is not mentioned on the AI Access Security or Prisma AIRS pages we read.
Read the profileZscaler
AI Access Security and Zscaler AI security (AI Gateway, AI Asset Management) on the Zero Trust Exchange
Zscaler states Coach and Block; Redact is partly stated; Watch is not in our sources.
States 6 of 13 checks, partly 3
Strength: States MCP and A2A brokers for agent traffic, plus discovery of models, agents and MCP servers.
Limit: Compliance certifications, including FedRAMP, are not stated on the pages we read.
Read the profile
Which pairs can I compare?
- Cato Networks and Cisco
- Cato Networks and Fortinet
- Cato Networks and Netskope
- Cato Networks and Palo Alto Networks
- Cato Networks and Zscaler
- Cisco and Fortinet
- Cisco and Netskope
- Cisco and Palo Alto Networks
- Cisco and Zscaler
- Fortinet and Netskope
- Fortinet and Palo Alto Networks
- Fortinet and Zscaler
- Netskope and Palo Alto Networks
- Netskope and Zscaler
- Palo Alto Networks and Zscaler
Checks at a glance
| Check | Cato Networks | Cisco | Fortinet | Netskope | Palo Alto Networks | Zscaler |
|---|---|---|---|---|---|---|
| Finds shadow AI use | Stated | Stated | Stated | Partly stated | Stated | Stated |
| Inspects prompts for sensitive data | Stated | Partly stated | Stated | Stated | Stated | Stated |
| Controls AI agent and MCP traffic | Stated | Partly stated | Stated | Stated | Partly stated | Stated |
| Covers coding agents | Stated | Not in our sources | Not in our sources | Partly stated | Stated | Partly stated |
| Protects homegrown AI apps at runtime | Stated | Stated | Stated | Stated | Stated | Stated |
| Blocks inline in real time | Stated | Stated | Stated | Stated | Stated | Stated |
| Coaches or warns the user | Not in our sources | Partly stated | Stated | Partly stated | Stated | Stated |
| Has a watch or log-only action | Stated | Partly stated | Stated | Not in our sources | Not in our sources | Not in our sources |
| Redacts or anonymizes data in prompts | Stated | Not in our sources | Not in our sources | Not in our sources | Not in our sources | Partly stated |
| Can run on your own infrastructure | Partly stated | Not in our sources | Stated | Not in our sources | Not in our sources | Not in our sources |
| States FedRAMP | Not in our sources | Partly stated | Not in our sources | Partly stated | Partly stated | Not in our sources |
| Lists security attestations | Stated | Not in our sources | Partly stated | Stated | Stated | Not in our sources |
| One console for AI and network policy | Stated | Stated | Partly stated | Partly stated | Partly stated | Partly stated |
Questions
Why these six vendors?
Each sells network security from the cloud (SASE or SSE) and publishes AI security capabilities. This edition covers these six only.
Why does every profile list limits?
Because every product has gaps on its public pages, and a buyer needs to know them. A limit here means the vendor's pages we read do not state something, not that the product lacks it.
Do any vendors publish prices?
No. None of the six publishes list prices for these products on the pages we read.