Humane Intelligence

Humane Intelligence is a 501(c)(3) nonprofit focused on evaluating AI models and systems for accountability, responsibility, and fairness, with a stated mission of reducing barriers to deploying AI for social good. It offers three main program lines: AI red teaming, a semi-structured testing approach used to identify vulnerabilities and limitations in AI models and systems; AI contextual evaluations, described as mixed-method, bespoke assessments of a model or system's performance within a defined problem space; and bias bounties, collaboratively designed challenge sets that convene researchers, impacted communities, and domain experts to examine AI/ML systems, models, and datasets.

The organization also operates a web application built for the data collection components of its hosted red teaming events and contextual evaluation engagements, and has stated plans to refactor and release the application under an open source software license in 2026. Additional activities include policy work, working groups (including an AI in Public Health working group), fellowships, volunteering, and events.

In 2026 the organization has said it is developing an ontology/knowledge graph methodology for contextual AI evaluation, migrating its bias bounty program onto Zindi, a global data science challenge platform, after closing its final self-hosted bias bounty in November 2025, and publishing funding requests to support operational and programmatic goals. [0][1]

Founding story

Co-founder Dr. Rumman Chowdhury developed the concept of "bias bounties" while working at Twitter, and the organization unofficially launched in 2022 under the name "Bias Buccaneers" before being incorporated as a nonprofit in May 2023. [1]

Business model

Nonprofit 501(c)(3) structure combining philanthropic funding with paid client services. Red teaming events and contextual evaluations are offered as paid services, with pricing options available on request, and the organization publishes funding requests to support programmatic and operational goals. [0]

Paid client services for AI red teaming events (run on the organization's own software) and contextual evaluations, alongside funder support solicited through published funding requests. [0]

Traction

The organization reports more than 5,300 participants in its in-person and remote events across five continents, more than 30 completed red teaming, bias bounty, and contextual evaluations, and stakeholders and participants from more than 55 countries. [0]

Latest developments

As of 2026, Humane Intelligence is developing an ontology/knowledge graph contextual AI evaluation methodology, moving its bias bounty program onto the Zindi data science challenge platform after closing its final self-hosted bias bounty in November 2025, preparing an open source release of its AI red teaming application, and building out the AI in Public Health working group. Mala Kumar was named Executive Director in April 2026. Recent published insights include posts on the relationship between AI safety and AI evaluations (August 2026) and human-centered AI evaluation content for LinkedIn Learning (June 2026). [0][1]