We support research that keeps advanced AI safe, aligned, and in sync with the values of humanity. Our mission is to help researchers achieve reliable safety methods before AI is no longer under human control.

AI is advancing faster than our ability to test, verify, and control it. These systems will soon influence national security, infrastructure, finance, medicine, and public policy. A poorly aligned system might pursue narrow goals with destructive efficiency, similar to the paperclip maximizer. If powerful AI behaves in unexpected or deceptive ways, the consequences will be irreversible.