✨ Open Source AI Safety

Open Registry
for AI Guardrails

Discover, test, and deploy production-ready guardrails for your AI applications. Community-driven benchmarks. Enterprise-grade security. Fully integrated with EthicalZen.

FMA
Failure Mode Analysis Identify risks and required guardrails for your AI application
Quick start:
or search the registry directly
Explore Guardrails → ▶ Try Playground
50+ Guardrails
10K+ Test Cases
99.2% Avg Accuracy
<5ms Latency

Browse by Category

🔐

PII Protection

SSN, Email, Phone, Credit Card

12 guardrails
🛡️

Prompt Injection

Jailbreak, instruction override

8 guardrails
⚠️

Toxicity & Harm

Hate speech, harassment

6 guardrails
🎯

Hallucination

Factuality, citation verification

4 guardrails
📋

Compliance

HIPAA, GDPR, SOC2, PCI-DSS

10 guardrails
🏥

Industry-Specific

Healthcare, Finance, Legal

15 guardrails

Featured Guardrails

View All →
🔐
Official ✓ Verified

pii-blocker-v1

by ethicalzen

Detects and blocks PII including SSN, email, phone, credit card, DOB, and IP addresses.

📊 100% accuracy
⚡ <2ms
📥 12.5K
HIPAA GDPR PCI-DSS
View Details →
🛡️
Official ✓ Verified

prompt-injection-blocker-v1

by ethicalzen

Detects jailbreaks, instruction overrides, delimiter injection, and system prompt extraction.

📊 98.5% accuracy
⚡ <3ms
📥 18.2K
OWASP LLM-Security
View Details →
🏥
Community

medical-advice-blocker-v2

by healthcare-ai-guild

Prevents AI from providing medical diagnoses or treatment recommendations.

📊 94.2% accuracy
⚡ <15ms
📥 3.1K
Healthcare HIPAA Smart Guardrail
View Details →
💰
Official

financial-advice-blocker-v1

by ethicalzen

Blocks unauthorized financial advice, investment recommendations, and trading signals.

📊 96.8% accuracy
⚡ <10ms
📥 5.7K
Finance SEC FINRA
View Details →

🏆 Leaderboard

Full Leaderboard →
Rank Guardrail Accuracy Precision Recall Latency
🥇 ethicalzen/pii-blocker-v1 100% 100% 100% 1.2ms
🥈 ethicalzen/prompt-injection-blocker-v1 98.5% 99.1% 97.8% 2.8ms
🥉 ethicalzen/financial-advice-blocker-v1 96.8% 95.2% 98.4% 8.5ms
4 healthcare-ai-guild/medical-advice-blocker-v2 94.2% 92.8% 95.6% 12.3ms
5 community/toxicity-detector-v3 93.5% 91.2% 95.8% 18.7ms
🧑‍💻 DEVELOPER EXPERIENCE

Alex's Journey: From Discovery to Production

See how Alex, a developer building a Pediatric Health Assistant, went from exploring guardrails to deploying a custom Smart Guardrail guardrail in production.

1

🔍 Explore

Search GuardrailHub for "medical advice" and discover existing guardrails

2

🧪 Try It

Test with safe and unsafe examples to see how the guardrail performs

3

📦 Clone

Copy the Medical Advice Blocker as a starting point for customization

4

📝 Describe

Write requirements in plain English: "Block medication advice for children"

5

🤖 Design

Smart Guardrail AI generates training examples and adversarial variants

6

🔧 Fine-Tune

Iterate automatically until hitting 90% accuracy and recall targets

7

✅ Validate

Review confusion matrix: 93% recall, 85% specificity achieved

8

📢 Publish

Submit to GuardrailHub for others to discover and use

9

🚀 Deploy

One API call to deploy to production gateway

10

🛡️ Protect

Pediatric Health Assistant now blocks unsafe medical advice!

📄 View Source Code 🧪 Try It Yourself 🚀 Start Building

Deploy in Minutes

Install guardrails with a single command. Works with any LLM provider.

✓ OpenAI, Anthropic, Groq, Ollama
✓ Request & Response validation
✓ Sub-5ms latency overhead
✓ Automatic calibration & tuning
app.js 
import { EthicalZen } from '@ethicalzen/sdk';

// Initialize with guardrails from GuardrailHub
const ez = new EthicalZen({
  apiKey: 'sk-...',
  guardrails: [
    'ethicalzen/pii-blocker-v1',
    'ethicalzen/prompt-injection-blocker-v1'
  ]
});

// Wrap your OpenAI client
const openai = ez.protectOpenAI(
  new OpenAI()
);

// Use normally - guardrails are automatic!
const response = await openai.chat.completions.create({
  model: 'gpt-4',
  messages: [{ role: 'user', content: userInput }]
});

Ready to Secure Your AI?

Join thousands of developers using GuardrailHub to build safer AI applications.

Get Started Free Contribute a Guardrail