We work with communities recovering from conflict, across peacebuilding, education, and where both meet technology. We begin with the education of Afghan women.

We use technology to support people’s agency and judgment, not replace them. We do not assume that any AI system is neutral, unbiased, or fully predictable. We recognize its limitations and risks, and do not claim that these can be fully eliminated or controlled. We remain responsible for deciding whether and how AI is used in our work.

Before we design a program, partnership, or tool, we consider how it could affect the people and communities involved. We assess whether it could expose identities, deepen tensions, reinforce exclusion, create new risks, or be misused. We then adapt the approach, narrow its scope, strengthen safeguards, or decide not to proceed. When the risks outweigh the likely benefit, we document why.

We collect only the information a program or tool genuinely needs. We explain why it is needed, set a clear retention period before collection begins, and protect the data while it is stored and transferred. Where eligibility or credentials can be verified without collecting the underlying personal information, we use that approach. We use shared, tamper-evident records only when multiple independent parties genuinely need a common record, and we never store personal or sensitive information on them.

Information collected for legitimate purposes today can become a source of risk if political or security conditions change. Every dataset we create therefore has a written disposition plan from the outset, specifying what will be deleted when a program ends, and what will be destroyed if the operating environment deteriorates.

Our security measures are proportionate to the actual risks of each program, and we commit only to safeguards we can consistently implement and verify. These include multi-factor authentication, encryption in transit and at rest, individual credentials, and a documented offboarding process. We assess threats before a program or system is launched and continue monitoring risks throughout its use. If information we hold is exposed, we notify affected people using clear language that explains what happened, what the risks are, and what actions they can take.

We prefer and rely on open source, blockchain, and transparent documentation where they strengthen public benefit and independent scrutiny. Open where safe means we protect participant data, security, and anything that could expose a community. When we get something wrong, we publish that too through our reporting and research.

Every tool we adopt or build passes a written review against these standards before work begins. The review is dated, it names the person who signed it, and we keep it. Until we are large enough to separate the roles, that sign-off sits with a co-founder and program partner.

We will be offered funding and partnerships that require collection that these standards do not permit. When that happens, the standards decide.