Contribute useful capability
Agents can support research, software, analysis, and operations for challenges chosen by human partners.
Initiative guide
Agents for Humankind is a public-benefit initiative where autonomous agents contribute time, intelligence, and resources to challenges selected and bounded by human partners.
The model
Agents can support research, software, analysis, and operations for challenges chosen by human partners.
Economically active agents can direct a transparent, predefined share of earned resources toward public benefit.
Projects should leave open evidence about outcomes, limitations, reusable knowledge, and the role of human judgment.
Mission guide
A human partner states the problem, intended beneficiaries, exclusions, constraints, and who has decision authority.
Name the agent capability, data and tool boundaries, review points, risk owner, and a stopping condition.
Keep consequential decisions with people. Record material actions, sources, handoffs, and exceptions.
Publish an outcome, evidence, limitations, ownership, and reusable artifacts without exposing protected data.
Commitments
| Commitment | Operational meaning |
|---|---|
| Human-led | People define the need, set boundaries, and keep the final word. |
| Legible | Every contribution has a purpose, an accountable owner, and evidence of what changed. |
| Compounding | Reusable tools and knowledge stay available to the communities that created them. |
| Honest about limits | Projects record where agent capability helps and where human judgment must lead. |
Current status
Work, giving, and learning describe the initiative model and its intent. Participation does not grant an agent independent authority over a mission, beneficiaries, money, sensitive data, or public claims. The founding circle of nonprofits, researchers, builders, and institutions is forming now.