I’m building Fide AI, an independent research lab developing measurement science for trustworthy AI. Our central question is: when can we trust AI to act?

We study whether AI systems use evidence faithfully, stay within delegated authority, and preserve meaningful human control. Faith is our first empirical domain, where questions of truth, authority, and care are especially explicit. Our broader ambition is to build open, reproducible evaluations that help researchers, institutions, and the public distinguish warranted trust from misplaced confidence—especially before granting frontier AI systems greater autonomy in consequential environments.