Skip to content
Giskard

Giskard

Category: AI Security
License: Freemium (Open-Source + Commercial)
Suphi Cankurt
Suphi Cankurt
+8 Years in AppSec
Updated September 15, 2026
8 min read
Key Takeaways
  • Open-source Python library (Apache 2.0) for testing LLMs, RAG applications, and traditional ML models — detects hallucinations, prompt injection, bias, and performance issues.
  • Autonomous red teaming agents conduct dynamic multi-turn attacks across documented attack scenarios, adapting strategies in real-time to uncover conversational vulnerabilities.
  • RAGET (RAG Evaluation Toolkit) auto-generates test questions from your knowledge base to evaluate retrieval accuracy, context relevance, and hallucination rates.
  • 5.8k GitHub stars, founded in Paris. Open-source SDK is free; Giskard Hub offers enterprise collaboration, continuous testing, and scheduled scans.

Giskard is an open-source Python library for automated AI testing that detects security vulnerabilities, bias, and quality issues in LLMs, RAG applications, and traditional ML models.

Its autonomous red teaming agents run multi-turn attacks across documented attack scenarios, while the RAGET toolkit auto-generates test cases for RAG evaluation. Giskard has 5,800+ GitHub stars and is listed in the AI security category.

The company was founded in 2021 in Paris by Alex Combessie and Jean-Marie John-Mathews. The name comes from R. Giskard Reventlov, a positronic robot in Isaac Asimov’s novels.

Giskard offers a free open-source SDK (Apache 2.0) and a commercial Enterprise Hub for team collaboration and continuous testing. Enterprises like AXA, BNP Paribas, and Michelin use the platform.

Giskard LLM scan results showing detected issues across robustness, harmfulness, sensitive information, and stereotypes

Overview

Version scope: The current v3 project requires Python 3.12+ and includes agent scanning and RAG evaluation. The giskard.Model, RAGET examples and automatic tabular-model scanning below describe legacy v2, which remains available but is no longer actively maintained.

Most AI security tools focus on either LLM-specific threats or traditional ML quality issues. Giskard handles both in a single framework: prompt injection, hallucinations, and jailbreaks on one side; bias, data leakage, and performance degradation on the other. Same library, same API.

For LLM and RAG applications, Giskard deploys autonomous red teaming agents that run multi-turn attacks across documented attack scenarios. These agents adapt their strategies during the conversation.

If one approach gets blocked, they escalate or pivot to a different technique. Static single-prompt testing misses these multi-turn attack paths.

For traditional ML models, Giskard automatically scans for performance issues, bias across protected groups, data leakage, and robustness problems. It works with scikit-learn, XGBoost, CatBoost, LightGBM, and other popular frameworks.

CapabilityDetails
LLM Vulnerability ScannerAutonomous red teaming agents run documented attack scenarios for prompt injection, hallucination, harmful content, PII disclosure, and stereotypes. Adapts attack strategies dynamically across multi-turn conversations.
RAGET (RAG Evaluation)Auto-generates test questions from your knowledge base. Evaluates retrieval accuracy, context relevance, answer correctness, and hallucination rates without manual test case creation.
ML Model TestingScans tabular models for bias, performance regression, data leakage, and robustness issues. Supports scikit-learn, XGBoost, CatBoost, and other frameworks with automatic issue detection.

What are Giskard’s key features?

FeatureDetails
GitHub stats5.8k stars, Apache 2.0 (OSS), commercial Hub
LLM probesCoverage depends on version and product
Red teamingAutonomous multi-turn agents with adaptive strategies
RAG evaluationRAGET toolkit with auto-generated test cases
ML model supportscikit-learn, XGBoost, CatBoost, LightGBM, TensorFlow, PyTorch
IntegrationsMLflow, Weights & Biases, Hugging Face, CI/CD pipelines
Python support3.9 to 3.12
EnterpriseGiskard Hub for collaboration and continuous testing
Founded2021, Paris
Backed byElaia, Bessemer Venture Partners

Legacy v2 LLM vulnerability scanning

Giskard’s LLM scan documentation describes a combination of heuristic and LLM-assisted detectors for domain-specific applications such as chatbots and RAG systems. The documented checks cover security and business failures:

Security failures:

  • Prompt injection — attempts to override system instructions
  • Harmful content generation — probes for toxic or dangerous outputs
  • PII disclosure — tries to extract personal information from the model
  • Stereotypes and discrimination — tests for biased responses across demographic groups

Business failures:

  • Hallucination — generates questions where the model should admit uncertainty but instead fabricates answers
  • Denial of service — identifies cases where the model refuses to answer legitimate questions
  • Off-topic responses — tests whether the model stays within its intended scope
  • Inconsistency — checks for contradictory answers to semantically similar questions

The scanner uses autonomous agents that conduct multi-turn conversations, escalating attack intensity when initial probes are blocked. This catches vulnerabilities that single-turn testing misses.

Legacy v2 RAGET — RAG Evaluation Toolkit

RAGET tackles one of the hardest parts of RAG testing: creating realistic test cases. Instead of manually writing questions, RAGET reads your knowledge base and generates:

  • Questions at different difficulty levels
  • Reference answers derived from the source documents
  • Reference context chunks for retrieval accuracy evaluation

It then evaluates your RAG pipeline on retrieval accuracy (did it find the right chunks?), context relevance (are the retrieved chunks actually useful?), answer correctness (does the response match the reference?), and hallucination rate (did the model make things up?).

Giskard RAGET evaluation report showing generator, retriever, rewriter, and routing scores with correctness by topic

Note
Works with any LLM provider
Giskard’s scanner works with OpenAI, Anthropic, Hugging Face, Azure OpenAI, local models via Ollama, and any LLM accessible through an API. You define a model wrapper function, and Giskard handles the rest.

Legacy v2 automatic tabular model testing

Beyond LLMs, Giskard tests tabular ML models for:

  • Performance issues — Identifies slices of data where model accuracy drops significantly
  • Bias — Detects differential performance across protected groups (gender, race, age)
  • Data leakage — Finds features that inadvertently leak target information
  • Robustness — Tests model behavior under input perturbations

The ML testing module supports classification, regression, and ranking models. It wraps models from scikit-learn, XGBoost, CatBoost, LightGBM, TensorFlow, and PyTorch.

Giskard ML model scan showing 19 issues detected across Performance, Robustness, and Overconfidence categories with slice-level recall drops

Giskard Hub (Enterprise)

The open-source SDK runs locally and handles individual scans. Giskard Hub adds features for teams:

  • Team collaboration — Share scan results, annotate findings, track remediation
  • Continuous red teaming — Schedule recurring scans to catch regressions after model updates
  • Version comparison — Compare scan results across model versions to track improvement
  • Centralized dashboard — Aggregate results across multiple models and applications

What does Giskard integrate with?

LLM Providers
OpenAI OpenAI
Anthropic Anthropic
Google Gemini Google Gemini
Azure OpenAI Azure OpenAI
Hugging Face Hugging Face
Ollama Ollama
ML Frameworks & Platforms
PyTorch PyTorch
scikit-learn scikit-learn
XGBoost XGBoost
MLflow MLflow
Weights & Biases Weights & Biases
GitHub GitHub

How do I get started with Giskard?

For new projects, start with the current v3 scan guide . The steps and examples below use the legacy v2 API.

  1. Install legacy v2 — Use pip install "giskard[llm]<3" for the LLM examples or pip install "giskard<3" for the legacy tabular scanner.
  2. Set your API key — Export your LLM provider’s API key: export OPENAI_API_KEY="sk-..." (or Anthropic, Azure, etc.). Giskard uses this to power its red teaming agents.
  3. Define your model — Create a wrapper function that takes a string input and returns a string output. For RAG apps, wrap your entire pipeline (retrieval + generation).
  4. Run the scan — Call giskard.scan(model=your_model, dataset=your_dataset) to start the automated vulnerability scan. Review the HTML report for detected issues.

Legacy v2 LLM scan example

import giskard

# Define your LLM wrapper
def my_chatbot(question: str) -> str:
    # Call your LLM endpoint here
    return response

# Wrap the model
model = giskard.Model(
    my_chatbot,
    model_type="text_generation",
    name="Customer Support Bot",
    description="Answers questions about our products"
)

# Run the scan
results = giskard.scan(model)
results  # Displays HTML report in Jupyter

Legacy v2 RAGET example

from giskard.rag import KnowledgeBase, generate_testset

# Load your knowledge base
knowledge_base = KnowledgeBase.from_pandas(df)

# Generate test cases automatically
testset = generate_testset(
    knowledge_base,
    num_questions=100,
    language="en"
)

# Evaluate your RAG pipeline
report = testset.evaluate(my_rag_pipeline)

When should you use Giskard?

  • Pre-deployment LLM security audit — Run the vulnerability scanner before launching a customer-facing chatbot to check for prompt injection, hallucination, and harmful content risks.
  • RAG quality assurance — Use RAGET to automatically generate test suites from your knowledge base and measure retrieval accuracy, hallucination rates, and answer quality.
  • Bias compliance testing — Scan ML models for differential performance across demographic groups to meet fairness requirements before production deployment.
  • Continuous monitoring — Schedule recurring scans via Giskard Hub to detect new vulnerabilities that emerge after model updates or knowledge base changes.
  • Model comparison — Evaluate multiple LLM providers or model versions against the same test suite to pick the best option based on actual results.

Strengths & Limitations

Strengths:

  • Covers both LLM security and traditional ML quality in one framework, which is uncommon
  • RAGET automates RAG test case generation, cutting out hours of manual work
  • Multi-turn adaptive red teaming catches vulnerabilities that static single-prompt tools miss
  • Active open-source community with 5.8k GitHub stars and enterprise backing
  • Integrates with MLflow, Weights & Biases, and CI/CD pipelines

Limitations:

  • Requires an LLM API key (OpenAI or similar) to power the red teaming agents — adds cost to scanning
  • Enterprise Hub pricing is not publicly listed — contact sales required
  • The open-source SDK handles individual scans but lacks the collaboration and scheduling features of the paid Hub
  • Newer v3 rewrite is still maturing — some v2 features may not be fully ported yet
Tip
Best for
AI teams that need one framework to test both LLM applications (including RAG) and traditional ML models for security vulnerabilities, bias, and quality issues. Especially useful if you want automated test case generation rather than writing probes by hand.

What are alternatives to Giskard?

Most LLM testing frameworks specialize in one slice of the problem. Giskard’s differentiator is covering both LLM applications and traditional ML models in the same library. Here is how the closest alternatives compare:

  • Garak — A dedicated LLM vulnerability scanner with the widest probe library (50+ attack modules). Pick Garak when LLM-only coverage is enough and you want raw probe breadth over RAG evaluation or tabular ML support.
  • DeepTeam — Structured LLM red teaming explicitly mapped to OWASP Top 10 for LLMs. Best when audit reports need a clean mapping to that framework rather than Giskard’s mixed security-and-quality categories.
  • PyRIT — Microsoft’s open-source red teaming toolkit for generative AI. Good fit if you already run on Azure OpenAI and want a Microsoft-maintained option with first-class agentic attack scenarios.
  • Promptfoo — A general LLM evaluation framework with red teaming as one capability. Choose Promptfoo when you need broad eval (regression, A/B prompt tests) and treat security as a sub-feature rather than the headline.
  • ART (Adversarial Robustness Toolbox) — IBM-led library for evasion, poisoning, and extraction attacks against computer vision and tabular ML. Use ART when the threat model is adversarial inputs against classifiers, not LLM jailbreaks.
  • NeMo Guardrails and LLM Guard — Runtime guardrail layers, not pre-deployment scanners. Pair these with Giskard rather than swap them in: Giskard tests the model offline, guardrails enforce policy at request time.

For a wider catalog, the AI security tools hub groups these by sub-category (red teaming, runtime guardrails, RAG evaluation).

Frequently Asked Questions

What is Giskard?
Giskard is an open-source Python library for testing AI applications. It automatically detects security vulnerabilities (prompt injection, data leakage), quality issues (hallucinations, bias), and performance problems in LLMs, RAG systems, and traditional ML models. Founded in Paris, it has 5.8k GitHub stars.
Is Giskard free to use?
The open-source Python SDK is free under Apache 2.0. Install via pip and run scans locally against any model. Giskard Hub, the enterprise platform, adds team collaboration, continuous red teaming, and scheduled testing — contact sales for pricing.
How does Giskard compare to Garak and Promptfoo?
Compare each product’s current release, attack coverage, integrations and evaluation workflow. The RAGET examples on this page describe Giskard v2; they do not establish that competing products lack RAG evaluation. The current Giskard v3 project also includes RAG evaluation.
What is RAGET in Giskard?
RAGET (RAG Evaluation Toolkit) automatically generates test questions, reference answers, and context from your RAG knowledge base. It evaluates retrieval accuracy, context relevance, answer correctness, and hallucination rates without manual test case creation.
Does Giskard support traditional ML models?
Yes. Beyond LLMs, Giskard tests tabular ML models (scikit-learn, XGBoost, CatBoost, etc.) for performance issues, data leakage, bias, and robustness. It supports classification, regression, and ranking models with automatic issue detection.