Meta's oversight board warns: committees drawn from your own boards could become pure window dressing
Six years of independent oversight of Meta has taught the board's members one thing: a watchdog without real power is not oversight, but decoration. Now, in an open letter responding to the White House's voluntary superintelligence accord, they are warning AI companies against designing their safety committees with the same structural flaws the board itself has experienced.
Members of Meta's Oversight Board have a message for AI companies: give your safety committees real power, or they mean little. In interviews, the members say the independent safety committees that AI companies including Google, Anthropic, OpenAI, Meta, xAI and Nvidia pledged to create last week could end up as little more than window dressing — unless they are given a clear mandate, genuine independence and the ability to push back against management. The warning comes in the form of an open letter, sent in response to the White House's voluntary Accord on Super Intelligence, and draws on the board's own experience of six years ruling on Meta's content moderation (MSN).
A voluntary accord without bindings
The framework that prompted the letter is, then, the White House's Accord on Super Intelligence. According to the reporting, the accord binds none of the six companies — OpenAI, Anthropic, Google, Meta, Nvidia and xAI — to any particular action. And that is a detail that weighs heavily in the board's warning: the Accord specifies that the intended independent committees that will oversee the companies' safety work are to consist of members drawn from the companies' own boards of directors.
That stands in sharp contrast to how the Oversight Board itself was constructed. The board includes no members of Meta's board, but is staffed by independent experts — among them journalists, scholars of human rights law, a Nobel Peace Prize laureate and a former Danish prime minister. And although Meta has funded the board with more than $300 million since 2019, the funds are held in an independent trust, separate from the company's own coffers.
The board's recipe: six recommendations
In the open letter, the Oversight Board sets out six recommendations for how independent bodies should exercise oversight of AI companies. According to the reporting, they concern establishing a concrete mandate, ensuring structural stability and independence, involving cross-disciplinary expertise, and emphasizing transparency.
The letter also argues for a broader governance model: "With a tool that has as much potential and as much risk as AI, we need a layered approach to governance that encompasses companies' own policies, industry-adopted standards, independent oversight of those standards from a range of perspectives, and both regulation and international coordination," the group writes, according to the reporting.
The members' concrete advice points toward the same principles. Paolo Carozza, a board member and professor of law and political science at the University of Notre Dame, points to friction as the very test: "If it's going to be genuine oversight, there will be friction between the company and the body," he says of the board's recommendations for the AI companies' oversight arrangements.
Nighat Dad, a board member and executive director of the Pakistan-focused Digital Rights Foundation, stresses two requirements: the committees must have real access to non-public information, and they must be financially independent.
Julie Owono, also a board member, points to transparency as a precondition for public trust: "Transparency is paramount," she says.
Why now
The timing is no accident. The board is speaking out before the AI companies have fixed the shape of their oversight arrangements. Carozza admits the board itself could have been more effective had it been constructed differently — and that it is precisely this hindsight that has driven it to share its experience now, before the companies cement their own plans.
The experiential basis is real, though limited. For six years, the board has ruled on the company's content moderation decisions. Even this arrangement has had weaknesses: Meta controlled part of the process for selecting the board's members until 2023 — a reminder that even well-intentioned oversight bodies can have structural flaws baked in from the start.
The test the committees will face
The warning can be turned on any prospective safety committee. Taking the board's criteria literally, the questions are concrete: Does the committee have a concrete mandate, or only a statement of purpose? Do the company's own board members sit on it, as the Accord envisions, or is it staffed by independent external experts, as the Oversight Board chose? Does it have financial independence from the company it is meant to check — the second of Dad's two requirements? Does it get access to non-public information — the first? And does it open itself to public scrutiny, as Owono insists?
The board's own history illustrates why the questions matter. The funding in an independent trust can be read as protection against Meta withdrawing support as punishment for unpopular decisions. The exclusion of Meta board members prevents the company from sitting in the room that judges it. The contrast with the Accord — where the committees are drawn from the companies' own boards and the agreement binds no one to action — is the underlying point of the entire letter.
A skeptical counterargument
The warning does not come without pushback. Evelyn Douek, a law professor at Stanford, believes caution is warranted in holding up the Oversight Board as a model. "I think the Oversight Board should be more of a cautionary tale than a sort of promising model of what independent oversight should look like in these situations," Douek says. Her argument points to the board's limited remit: it rules only on a selection of Meta's decisions, not on the company's operations as a whole.
Carozza appears to partially accept the point. That the board could have been constructed better is part of the message — the experience is valuable precisely because it includes the mistakes.
What we don't know
Several things remain unresolved. In this reporting, the companies have not offered their own statements on how they envision the promised committees, so their side of the story is so far undocumented. The exact terms of the Accord on Super Intelligence and the wording of the board's open letter are known through secondary reporting, not through the documents themselves. And when the arrangements are presented, the decisive question will be how many of the board's criteria they meet on paper — and whether the friction Carozza describes actually arises in practice.