Meta Oversight Board members warn AI safety panels need real authority
In new interviews after the voluntary White House accord, board members say AI oversight bodies need independence, access to internal information and power to require change.
A warning about oversight without power
Members of Meta’s Oversight Board told NBC News in interviews published October 7 that proposed AI safety committees could become symbolic unless they can challenge company management. Board member Paolo Carozza said effective independent oversight should create friction with the company. Nighat Dad argued that committees need access to nonpublic information and control over their own funding and membership. These are the members’ judgments about how oversight should work, not findings that the six companies in the White House Super Intelligence accord have already failed to meet a standard.
The older letter and the new interviews
The board’s underlying open letter was published September 30, outside this article’s immediate news window. It recommends transparent reasoning, human-rights standards, outside expertise, structural independence, a concrete mandate and industrywide coordination. NBC’s October 7 interviews add members’ explanations of why those design choices matter. The White House accord is voluntary and calls for committees within company boards, while Meta’s Oversight Board consists of outside members and has binding authority over a limited set of content decisions. The board itself acknowledges those limits. No new AI oversight body or binding regulation was created by the interviews or the letter.