What Giskard does
Giskard is a Paris-based AI testing company founded in 2021. Its open-source library and enterprise Giskard Hub help teams continuously red-team and test large language models and AI agents before deployment. It detects security vulnerabilities such as prompt injection, data disclosure and harmful content, alongside quality issues like hallucination, contradiction and omission, then ranks findings by severity with go/no-go recommendations.
Open source plus enterprise
The open-source core is freely available on GitHub and widely used by the ML community. The commercial Hub adds role-based access control, audit trails, EU and US data residency and on-premise deployment, with compliance covering GDPR, SOC 2 Type II and HIPAA. Giskard integrates with GitHub, major clouds, NVIDIA, Databricks, Mistral AI and Hugging Face.
Funding
Giskard has raised roughly USD 8M, including a seed round led by Elaia with Bessemer Venture Partners and Y Combinator, plus a European Innovation Council grant and a Bpifrance research grant supporting LLM evaluation research.