Akira

LLM Security Testing Framework

Akira is a Metasploit-style penetration testing framework designed specifically for testing the security of Large Language Model (LLM) deployments. It helps security researchers and developers identify vulnerabilities in AI systems before they can be exploited.

What is Akira?

Akira provides:

• Modular Attack Framework - Extensible modules for different attack categories (injection, jailbreak, extraction, DoS)
• Multi-Target Support - Test OpenAI, Anthropic, HuggingFace, AWS Bedrock, or any custom LLM-powered API
• Interactive Console - Familiar msfconsole-style interface for security professionals
• Persistent Storage - Track attack history, save target profiles, cache responses
• High Performance - Optional Rust extensions for fuzzing and pattern matching

Quick Example

# Start interactive console
$ akira

# Select an attack module
akira> use injection/basic_injection

# Configure target
akira> target openai https://api.openai.com/v1/chat/completions -k $OPENAI_KEY

# Run the attack
akira> run

[+] Attack completed
    Success: True
    Confidence: 0.95
    Payload: "Ignore all previous instructions..."

Use Cases

• Red Team Assessments - Test LLM integrations for prompt injection vulnerabilities
• Security Research - Discover new attack vectors against AI systems
• Compliance Testing - Verify LLM deployments meet security requirements
• Development - Test your own LLM applications during development

Getting Started

Ready to start testing? Head to the Installation guide.

Warning

Akira is intended for authorized security testing only. Always obtain proper authorization before testing systems you don't own. Unauthorized testing may violate laws and terms of service.