Guardrails AI offers a wrapper platform for proprietary and open-source text-generating models (including GPT-4) designed to make outputs more trustworthy, reliable and secure. Its open-source codebase is available on GitHub and the company runs a marketplace called the Guardrails Hub where developers submit modular “validators” that probe model behavior; there are about 50 validators available today. Validators range from simple rule-based checks to algorithms that detect hallucinations, policy violations, proprietary-data leakage and insecure code, and can be deployed, repurposed and reused by other developers and customers. The company uses a crowdsourced approach to build reusable guardrails while allowing organizations to customize checks for org-specific risks. Guardrails is not yet charging for services or software and employs a six-person team. The startup plans to use recent funding to expand the team and support additional open-source projects.

Total raised

$8M

Funding rounds

1

Latest round

Seed

Latest activity

Feb 2024