Giskard AI features, pricing and GDPR (October 2026)

Giskard tests conversational AI agents for security and quality problems before and after launch, with the enterprise Giskard Hub for continuous red teaming and an open-source library for developers.

Giskard Hub runs black-box tests against agents reachable by API and detects hallucinations, prompt injection, data disclosure and sycophancy, with more than 50 automated probes including multi-turn attacks, go/no-go readiness verdicts and remediation support. RAG evaluation, CI/CD integration, SSO and role-based access are included. It can be hosted in the EU or US or installed on-premises. The open-source library giskard-oss (Apache 2.0, version 3) offers checks and scans for developers. Giskard is based in Paris and states SOC 2 Type II.

• Red-teams agents continuouslyAutomated probes, including multi-turn attacks, test an agent repeatedly as it changes. Security teams see new weaknesses after each update instead of only at launch.
• Gives a release verdictProduction readiness assessments summarize findings into a go or no-go verdict. Product owners get a clear basis for release decisions.
• Evaluates RAG answersRAG evaluation measures answer quality against the knowledge base. Teams find where the assistant invents facts or misses relevant documents.

• French provider with EU hosting option and on-premises installation.
• Open-source library under Apache 2.0 alongside the enterprise Hub.

Giskard is based in Paris and names French customers such as Michelin, BNP Paribas and Decathlon.

Choose EU hosting or on-premises installation if test conversations contain personal data, and sign a data processing agreement. Giskard Hub is priced on request. Tests call your agent through its API, which creates load and model costs on your side.

No reviews yet. Be the first to share how your team uses it.

Every review shows its verification level. We read each one before it goes live. How we check reviews