The process of independent model evaluation sits at the center of AI governance. A federal testing apparatus is supposed to assess capability and safety properties across frontier models, providing the signal that policymakers need to set rules with some grounding in what the technology actually does. OpenAI has appointed a senior US official from that testing apparatus to its board, a move that has drawn accusations of regulatory capture against the major AI laboratories.
The revolving door
Model testing agencies occupy a specific position in the AI governance stack. They set evaluation criteria and measure safety and capability claims against those criteria. Critically, they are supposed to operate at arm's length from the labs whose models they assess. The value of their output depends entirely on that separation. When a senior figure from such an agency takes a board seat at one of those labs, the arm's-length principle starts to look more like a revolving door.
The appointment adds a layer of complexity to the relationship between OpenAI and the government bodies tasked with overseeing the field. OpenAI makes ChatGPT and is among the most prominent labs in the frontier model space. Placing a former testing official in a board seat is a different order of engagement from policy advocacy.
The personal dimension has amplified scrutiny. The incoming board member previously lived with the chief of Anthropic, OpenAI's primary rival. That biographical detail, surfaced alongside the appointment, reinforces concern about the density of the network connecting government AI oversight to the labs themselves.
The regulatory capture charge
Regulatory capture describes the process by which a regulated industry gains undue influence over the agencies meant to police it. AI critics have applied that frame to a pattern in which lab-aligned figures move into advisory bodies, safety institutes, and corporate boards that shape how models are built, evaluated, and governed. OpenAI's board appointment is being read as a data point in that pattern, not an anomaly.
The criticism lands at a moment when the credibility of independent AI evaluation is already being questioned and the testing infrastructure itself is still being built. The constraint is institutional independence. A board seat at one of the labs is a direct claim against it.