To break down barriers to entry to AI evaluations, Humane Intelligence is working to release our red teaming web application under an open source software (OSS) license. Originally conceived as a software to allow anyone to red team commonly used AI models, our OSS release will improve our software, and make it easier and cheaper to meet the highest demand to use our software, which is among organizations that are building AI for good models and systems.
Our goal is to end with a robust OSS release + improve our managed instance, to give organizations wishing to red team their AI systems a broad range of licensing and engagement options to fit a spectrum of budgets and needs. To complete this work, we have established a simple forking strategy below. Our work kicked off in July 2025, with plans to complete phase 3 by February 2026.
Based on recommendations from Tattle Civic Tech, we plan to release our AI red teaming app under the “Trusted Vendor” OSS archetype. This will allow Humane Intelligence to continue to run our managed instance for organizations that do not wish to work with code. Other organizations can use the OSS release and are not locked into using us as a vendor.
As part of their investigation of our current code base, Tattle Civic Tech considered three possible options for the AI red teaming app to be modified ahead of its OSS release. The preferred option – Option C – is to: