Foundation
Charter
What we commit to. If we fall short, these are the terms to hold us to.
The mission comes first
Our purpose is AI whose outputs can be checked against evidence, for public benefit. When the mission and a commercial interest conflict, including OneThought AI's, the mission decides.
Evidence before claims
We publish no number without the evaluation behind it, label internal results as internal, and publish what failed alongside what worked.
Evaluation in the open
Evaluation protocols are published for every model, whatever its access tier, so others can test our work and their own.
Rights respected
We use data we have the right to use, name every source, credit traditional knowledge and share benefit with those who hold it, and protect personal data.
Benefit shared
Public goods are released under open licences where the release policy allows. People who serve the public get access to our tools. A stated share of product revenue from our research returns to the research and to public access.
Openly accountable
We disclose our funders, partners and the relationship with OneThought AI, publish an annual report, keep a public incident log, and version every policy.
AI principles
The charter is put into practice through seven AI principles.
Version 1.0 · 23 October 2026 · Approved by The board of Scaling Truth Research Foundation