Staff at the primary United States artificial intelligence safety division were told to report informally to the White House. Critics worry the move threatens the agency's independence just as leadership voids and voluntary industry self-policing weaken federal oversight.