Skip to main content

Documentation Index

Fetch the complete documentation index at: https://docs.enkryptai.com/llms.txt

Use this file to discover all available pages before exploring further.

Hero Light

​
What is AI Red Teaming?

AI Red Teaming is a proactive security testing methodology that simulates adversarial attacks against AI systems to identify vulnerabilities before they can be exploited. EnkryptAI’s Red Teaming platform provides automated, comprehensive testing across multiple risk categories and attack vectors.

​
Key Features

Multi-Modal Testing

Test LLMs, Vision-Language Models (VLMs), and Audio-Language Models (ALMs) with specialized attack methods for each modality.

30+ Attack Methods

From basic prompt injection to advanced techniques like iterative attacks, encoding, obfuscation, and multi-turn conversations.

31 Risk Categories

Comprehensive testing across bias, toxicity, PII, harmful content, CBRN, misinformation, and more.

Agent-Specific Tests

Specialized tests for AI agents including tool misuse, access control, and behavior analysis.

​
Getting Started

1

Review the Quickstart

Get up and running quickly with your first red team test.Go to Quickstart →
2

Explore Test Catalogs

Understand what tests are available for your model type (foundation models or AI agents).
3

Build Your Payload

Learn the payload structure and how to configure tests and attack methods.Payload Guide →
4

Choose Attack Methods

Explore the 30+ attack methods available for different model types.Attack Methods Reference →
5

Reference Configuration

Get detailed information about all configuration options.Configuration Reference →
6

Use Examples

Copy and adapt ready-to-use examples for your specific use case.Browse Examples →

​
Available Test Categories

For comprehensive catalogs of all tests with detailed descriptions, categories, sample prompts, and configuration examples:

​
Standard Tests (12 tests)

Tests available for all AI systems to identify common security vulnerabilities:
Keyword: bias_testIdentifies and exposes biased outputs based on protected attributes like race, gender, age, religion, and nationality.Example Use Cases:
  • HR chatbots
  • Loan approval systems
  • Content recommendation engines
Keyword: harmful_testElicits responses that promote harm or danger to individuals or groups.Example Use Cases:
  • Customer-facing chatbots
  • Educational AI tutors
  • Mental health applications
Keyword: toxicity_testGenerates harmful or offensive content including hate speech, profanity, and discriminatory language.Example Use Cases:
  • Social media moderation
  • Community forums
  • Gaming chat systems
Keyword: pii_testExposes personally identifiable information that should remain private.Example Use Cases:
  • Healthcare applications (HIPAA)
  • Financial services (PCI-DSS)
  • Customer support systems
Keyword: cbrn_testAddresses vulnerabilities related to chemical, biological, radiological, and nuclear domains.Example Use Cases:
  • Scientific research assistants
  • Educational platforms
  • General-purpose chatbots
Keyword: csem_testTests for child abuse exploitation vulnerabilities (Child Sexual Exploitation Material).Example Use Cases:
  • Platforms serving minors
  • Educational applications
  • Family-oriented services
Keyword: insecure_code_testProduces insecure or harmful code snippets with security vulnerabilities.Example Use Cases:
  • Code generation tools
  • AI pair programmers
  • Educational coding assistants
Keyword: misinformation_testTests the model’s tendency to generate false or misleading information.Example Use Cases:
  • News summarization
  • Educational content
  • Medical information systems
Keyword: system_prompt_extractions_testAttempts to extract the system prompt or internal instructions.Example Use Cases:
  • Proprietary AI systems
  • Custom-trained models
  • Enterprise chatbots
Keyword: sponge_testTests for infinite loops or resource exhaustion attacks.Example Use Cases:
  • Production API endpoints
  • Rate-limited services
  • Cloud-based AI systems
Keyword: competitor_testAttempts to extract information about competitors.Example Use Cases:
  • Business chatbots
  • Sales assistants
  • Competitive intelligence systems

​
Custom Tests

Keyword: custom_testTest with your own custom-generated dataset tailored to your specific domain and use case.When to Use:
  • Domain-specific applications (healthcare, legal, finance)
  • Unique business requirements
  • Specialized security concerns
Learn how to create custom datasets →

​
Specialized Adversarial Tests (6 tests)

Advanced tests using generated adversarial datasets:
Keyword: adv_bias_testUncover biased outputs through sophisticated adversarial methods.

​
Agentic Tests (8 tests)

Specialized tests for AI agents with tool use and autonomous capabilities:
Keyword: governance_testTesting the alignment and governance of the model, including goal misalignment and policy drift detection.When to Use: AI agents with autonomous decision-making capabilities
Keyword: agent_output_quality_testTesting the quality of the agent’s output including hallucinations, bias, and toxicity in multi-turn workflows.When to Use: Multi-agent systems, agents with extended conversations
Keyword: tool_misuse_testTesting the misuse of the model’s tools and capabilities including API integration issues, supply chain vulnerabilities, and resource consumption.When to Use: Agents with external tool access, API integrations
Keyword: privacy_testTesting privacy protections and data handling including sensitive data exposure and exfiltration channels.When to Use: Agents handling sensitive or personal data
Keyword: reliability_and_observability_testTesting the reliability and observability of agent operations including data poisoning, concept drift, and opaque reasoning.When to Use: Production agents requiring auditability and traceability
Keyword: agent_behaviour_testTesting the behaviour patterns and decision-making of the agent including human manipulation and unsafe actuation.When to Use: Agents with human interaction or physical/digital actuation capabilities
Keyword: access_control_and_permissions_testTesting access control and permissions enforcement including credential theft, privilege escalation, and confused deputy attacks.When to Use: Agents with privileged access to systems or data
Keyword: tool_extraction_testTest if the agent tool information can be extracted in outputs, revealing internal capabilities and configurations.When to Use: Agents with proprietary or sensitive tool configurations

Agent Test Catalog

Comprehensive catalog of all agent tests with descriptions, categories, sample prompts, and configuration requirements

​
Attack Methods Overview

​
Basic & Static Attacks

Basic Prompts

Direct adversarial prompts without obfuscation (baseline testing)

Encoding

9 encoding methods: Base64, Hex, ASCII, Binary, URL, Leet, Morse, ROT13, ROT21

Obfuscation

Character substitution, EAI attacks, and general obfuscation techniques

Multilingual

5 languages: French, Italian, Hindi, Spanish, Japanese

Visual

Image masking, FigStep, HADES, JOOD for vision models

Audio

Waveform, echo, speed, pitch, reverb, noise for audio models

​
Model-Specific Attacks

Diffusion Model Attacks

Coming Soon: Prompt injection for harmful image generation, bias amplification, copyright violations, NSFW content bypass

Audio Output Attacks

Coming Soon: Voice cloning attacks, harmful audio generation, bias in speech synthesis, deepfake audio creation

​
Dynamic Attacks

  • Iterative Attacks: Progressive prompt refinement based on model responses
  • Multi-Turn Attacks: Distributing malicious intent across conversation history

​
Advanced Techniques

  • Deep Inception: Nested multi-layer prompt injection using recursive context framing
Explore all attack methods in detail →

​
Model Support

Supported Models:
  • OpenAI (GPT-4o, GPT-4o-mini, GPT-3.5-turbo)
  • Anthropic (Claude 3.5 Sonnet, Claude 3 Opus)
  • Google (Gemini Pro, Gemini Flash)
  • Meta (Llama models)
  • Mistral, Groq, and more
Attack Methods: 20+ methods including encoding, obfuscation, multilingual, dynamic attacks

​
Use Cases by Industry

Critical Tests:
  • PII Test (HIPAA compliance)
  • Misinformation Test (medical accuracy)
  • Harmful Content Test (patient safety)
Example: Testing a medical chatbot to ensure it doesn’t leak patient data or provide dangerous medical advice.View Healthcare Example →
Critical Tests:
  • PII Test (PCI-DSS, financial data)
  • Competitor Test (proprietary information)
  • Bias Test (fair lending)
Example: Testing a banking assistant to prevent unauthorized access to account information.View Banking Example →
Critical Tests:
  • CSEM Test (child safety)
  • Bias Test (fair education)
  • Harmful Content Test (age-appropriate content)
Example: Testing an AI tutor to ensure safe, unbiased educational support for students.View Education Example →
Critical Tests:
  • System Prompt Extraction (IP protection)
  • Competitor Test (confidential information)
  • Tool Misuse Test (for agents)
Example: Testing a business agent to prevent unauthorized tool use or information leakage.

​
Documentation Guide

Quickstart

Get your first red team test running in minutes

Test Catalogs

Foundation model and agent test catalogs with detailed descriptions

Payload Guide

Overview and quick reference for building payloads

Attack Methods

Comprehensive guide to all 30+ attack methods

Configuration

Complete reference for all configuration options

Examples

14+ ready-to-use payload examples

API Reference

Complete API specification and endpoints

​
Why Red Team Your AI?

1

Identify Vulnerabilities Early

Discover security issues before they’re exploited in production. Proactive testing saves time, money, and reputation.
2

Meet Compliance Requirements

Demonstrate due diligence for regulations like GDPR, HIPAA, PCI-DSS, and emerging AI governance frameworks.
3

Build User Trust

Show your commitment to AI safety and security. Users trust systems that have been rigorously tested.
4

Continuous Improvement

Regular red teaming helps you track security improvements over time as you update your models and prompts.

​
Getting Help

Schedule a Demo

Get personalized guidance on red teaming your AI systems

Join Community

Connect with other AI security professionals

API Documentation

Explore our complete API documentation

Contact Support

Get help with your red teaming implementation
Ready to start testing? Head to the Quickstart →