Introduction to Axiom

Axiom is a runtime security scanner for AI and LLM applications. It evaluates application-controlled boundaries and returns a versioned verdict and action.

What is Axiom?

Axiom lets an application scan user input, retrieved context, proposed tool calls, and model output. Each deployment binds a versioned security profile to its API keys.

Core Capabilities

  • Boundary Scanning: Evaluate input, retrieval, tool, and output content
  • Threat Detection: Identify prompt injection, jailbreak, and data-exfiltration attempts
  • Policy Enforcement: Return monitor or enforce actions from the active deployment profile
  • Decision Evidence: Record reason codes, tier paths, and policy/model versions

Why Axiom?

AI and LLM applications face unique security challenges that traditional security tools aren't designed to handle. Axiom was built from the ground up to address these challenges:

AI-Specific Threats

  • Prompt injection attacks
  • Indirect instructions in retrieved content
  • Sensitive-data extraction attempts
  • Jailbreak attempts and model misuse

How It Works

Axiom integrates through authenticated HTTPS endpoints or the SDK:

  1. Create a Deployment: Activate a versioned security profile for an application or environment
  2. Generate a Key: Bind a one-time API credential to that deployment
  3. Scan Boundaries: Call Axiom before or after the corresponding application operation
  4. Use the Action: Log monitor decisions or stop requests when enforcement is enabled

Next Steps

Ready to get started? Here's what to do next: