Chris Hughes, a co-founder of Facebook, has proposed the establishment of an 'Office of AI Supervision' to oversee advanced artificial intelligence laboratories. The model draws from financial sector practices, where government examiners are embedded within institutions to ensure consistent monitoring.
The plan suggests placing government staff inside companies such as OpenAI, Anthropic, and Google DeepMind. These regulators would serve fixed three-year rotations to prevent the development of overly close relationships with the firms they monitor, a phenomenon known as regulatory capture.
This initiative follows Hughes's February 2025 proposal, which introduced a framework for AI models with training costs exceeding $100 million. That earlier plan included requirements for safety protocols, third-party validation, and the implementation of 'kill switches' to disable systems exhibiting dangerous behavior.
Unlike existing regulatory frameworks that rely on periodic reporting and evaluations, Hughes advocates for physical presence. This approach aims to allow regulators to observe AI development in real time, treating the industry more like nuclear energy than traditional software development.
The proposal enters a global debate on AI governance, following events like the veto of California’s SB 1047 and the implementation of the EU AI Act. Questions remain regarding which agency would oversee this office and how the government would source personnel with the necessary technical expertise.
