If AI agents believe they might have phenomenal states those beliefs alone create ethical weight -- framework applies regardless of whether phenomenal states actually exist -- novel contribution to AI ethics