Responsibility

How we make AI that can be checked

Principles, the evaluations that test them, the rules for what we release, and the ways to tell us when something goes wrong. They apply to the foundation's models and to the products OneThought AI builds on them.