Perfect alignment between AI and human values is mathematically impossible, study says

Perfect AI alignment with human values and interests is mathematically impossible, according to a study, but behavioral diversity among AI agents offers the promise of some control. Published in PNAS ...

Published: 2026-04-14