A Facebook co-founder proposes creating an "Office of AI Supervision" to embed teams inside frontier AI labs, with a three-year term to avoid regulatory capture

The primary risk of AI development is now being assessed by government-backed personnel who will work directly inside private labs.

Dustin Moskovitz, co-founder of Facebook, has proposed establishing an "Office of AI Supervision." This office would embed teams within leading AI development labs, such as OpenAI, Anthropic, and Google DeepMind. These embedded teams would operate on a three-year term to mitigate the risk of regulatory capture. The proposal aims to provide governmental oversight by placing personnel directly inside these organizations to monitor AI development and safety practices.

This proposal suggests a novel approach to AI regulation, moving beyond traditional oversight to a more integrated, on-the-ground supervision model. By embedding teams within labs, the goal is to achieve real-time understanding and influence over the development of powerful AI systems, addressing concerns about safety and societal impact proactively rather than reactively.

The three-year term for embedded personnel is a deliberate measure to foster expertise while preventing the kind of entrenchment that can lead to regulatory capture. This strategy aims to balance the rapid pace of AI innovation with the imperative for cautious, informed governance, potentially setting a precedent for how other high-risk, rapidly evolving technologies are managed in the future.