Lakera Guard

Lakera Guard is an enterprise AI security platform that protects LLM applications against prompt injection, jailbreaks, and data leakage in real-time.

Known for creating the famous Gandalf prompt injection game.

What is Lakera Guard?

Lakera is a Swiss AI security company that provides real-time protection for Large Language Model applications.

Their flagship product, Lakera Guard, acts as a security layer between users and LLMs to prevent attacks.

The company gained widespread recognition for creating Gandalf, an educational game that challenges players to extract a secret password from an AI through prompt injection.

Key Features

Prompt Injection Detection

Lakera Guard detects and blocks prompt injection attacks:

• Direct prompt injection
• Indirect prompt injection
• Jailbreak attempts
• System prompt extraction

Content Moderation

Filter harmful or inappropriate content:

• Toxicity detection
• Hate speech filtering
• Violence detection
• Custom content policies

Data Leakage Prevention

Prevent sensitive data exposure:

• PII detection and redaction
• Secrets detection
• Custom data patterns
• Input/output monitoring

How It Works

Lakera Guard integrates as a security layer:

User Input → Lakera Guard → LLM → Lakera Guard → Response
                  ↓                      ↓
            Block attacks          Filter outputs

Integration

API Integration

import lakera

# Initialize client
client = lakera.Client(api_key="your-api-key")

# Check user input
result = client.guard.analyze(
    prompt="User input here",
    categories=["prompt_injection", "pii", "toxicity"]
)

if result.flagged:
    # Block the request
    return "Request blocked for security reasons"
else:
    # Proceed with LLM call
    response = call_llm(prompt)

Supported Platforms

Lakera integrates with:

• OpenAI (GPT-4, GPT-3.5)
• Anthropic (Claude)
• Google (Gemini, PaLM)
• Azure OpenAI
• AWS Bedrock
• Custom LLM deployments

Detection Categories

Gandalf Game

Lakera created Gandalf, an interactive game where players try to extract a secret password from an AI.

The game demonstrates prompt injection techniques and has been used by:

• Security researchers
• AI engineers
• Educational institutions
• CTF competitions

When to Use Lakera Guard

Lakera Guard is ideal for:

• Teams deploying LLM-powered applications
• Applications handling sensitive user data
• Customer-facing AI chatbots
• Compliance-sensitive environments