Giskard
Open source testing for LLM and ML apps, with a scan for common vulnerabilities.
About Giskard
The open-source product automatically runs exhaustive test suites to identify risks on LLMs. Giskard was founded in Europe by data scientists, AI engineers and AI researchers. The platform is also designed for business users such as domain experts and product managers, not only for developers.
Giskard was founded to address challenges in ensuring the performance, reliability and security of AI systems. Giskard launched the Phare LLM Benchmark in partnership with Google DeepMind to evaluate LLMs across key safety and security dimensions. For mission-critical workloads in the public sector, defense or other sensitive applications, the engineering team can help install Giskard Hub in on-premise environments.
Drafted from giskard.ai · not yet confirmed by Giskard
How it works
Giskard scans an LLM app for hallucination, harmful content, prompt injection, sensitive information disclosure, stereotypes and robustness problems, generating test cases automatically and producing a report, and it supports tabular ML models too. There's an open library and a hosted hub. French.
From the project
Reviews
Be the first to review it
AnonymousKnow a team weighing Giskard? Send them this page.