What Humane Intelligence does
Humane Intelligence is a U.S.-based 501(c)(3) nonprofit focused on making AI systems more accountable, responsible, and fair through participatory and adversarial evaluation. Its core approach combines (1) human-centered AI red teaming (adversarial and non-adversarial testing designed to surface vulnerabilities and limitations), (2) “bias bounties” that structure research challenges to identify and address exclusionary or biased AI outcomes, and (3) contextual (bespoke) AI evaluations intended to rigorously assess model behavior within a defined problem space. The organization also publishes evaluation playbooks and is building open-source tooling to reduce barriers to entry for third parties who want to run their own evaluations and public-interest testing. In 2025–2026, the organization emphasized open-source release of its AI red teaming web application under an Apache 2.0 license, and it has been expanding evaluation methodologies (including an ontology/knowledge-graph oriented approach to contextual evaluation) while convening global communities of practice around evaluation, transparency, and social impact.