Firewall for AI

cloudflare Web application firewallcloudflare-firewall-for-ai

WAF detection for LLM endpoints that flags PII, prompt injection and unsafe topics in prompts.

Provider: cloudflare
Kind: Web application firewall
Category: Security
Official docs: https://developers.cloudflare.com/waf/detections/firewall-for-ai/

Properties · 4 set by this component

property	type	values	required	kind default	this component
botProtection	boolean	—	no	—	▸ false
customRuleCount	integer	—	no	—	inherited
endpointLabel · component property	string	—	no	cf-llm	inherited
managed	boolean	—	no	false	▸ true
managedRuleSet	string	—	no	—	inherited
mode	enum	detection \| prevention \| off	no	—	▸ detection
piiDetection · component property	boolean	—	no	true	inherited
promptInjectionDetection · component property	boolean	—	no	true	inherited
rateLimiting	boolean	—	no	—	▸ false
unsafeTopicDetection · component property	boolean	—	no	true	inherited

Rules that apply · 0

Every rule whose selector covers this component's kind or its category. Whether a rule fires depends on the design: its properties, its edges and its groups.

No rule reads this kind yet. That is a gap, not a guarantee.

Use it in a design

nodes:
  - { id: "firewall-for-ai", component: "cloudflare-firewall-for-ai" }

Other Web application firewall products · 13

Web Application Firewallalibaba-wafalibabacloudwaf AWS WAF web ACLaws-wafawswaf AWS WAF Classicaws-waf-classicawswaf Azure Web Application Firewallazure-web-application-firewallazurewaf Cloudflare WAFcloudflare-wafcloudflarewaf Corazacorazawaf Imperva Application Securityimpervawaf ModSecuritymodsecuritywaf NAXSInaxsiwaf Generic web application firewallwafwaf Google Cloud Armorgoogle-cloud-armorgcpwaf reCAPTCHA Enterprisegoogle-recaptcha-enterprisegcpwaf Web Application Firewalloci-wafociwaf

Source

1# yaml-language-server: $schema=../../schema/component.schema.json
2# Firewall for AI (AI Security for Apps): WAF detection that scans prompts on endpoints
3# labelled cf-llm for PII, prompt injection and unsafe topics. It only populates detection
4# fields; blocking needs a custom or rate limiting rule ("Firewall for AI"), so it ships
5# in detection mode.
6schemaVersion: "1"
7type: "component"
8id: "cloudflare-firewall-for-ai"
9kind: "waf"
10provider: "cloudflare"
11docs: "https://developers.cloudflare.com/waf/detections/firewall-for-ai/"
12defaults:
13  managed: true
14  mode: "detection"
15  rateLimiting: false
16  botProtection: false
17properties:
18  promptInjectionDetection:
19    type: "boolean"
20    default: true
21  piiDetection:
22    type: "boolean"
23    default: true
24  unsafeTopicDetection:
25    type: "boolean"
26    default: true
27  endpointLabel:
28    type: "string"
29    default: "cf-llm"

The same page as markdown, for an agent: /catalogue/components/cloudflare-firewall-for-ai.md