Centre for the Governance of AI

The Centre for the Governance of AI (GovAI) describes itself as a think tank that aims to help decision-makers in government and industry navigate the transition to a world with advanced AI S-1818. It was founded at Yale University in 2016, was an academic centre at the University of Oxford from 2018, and became an independent nonprofit in 2021 S-1818. Technical governance is one of its focus areas S-1818.

Two of its policy papers bear on verification:

• A 2024 paper by Sastry, Heim and colleagues argues that AI-relevant compute is "detectable, excludable, and quantifiable" and produced through an extremely concentrated supply chain S-0053. It describes compute as a lever for regulatory visibility, for allocating resources and for enforcing restrictions on AI development and use S-0053. The claims on compute stock, inference versus training and bandwidth between compute groups draw on it.
• A 2023 report by Trager and colleagues proposes an International AI Organization that certifies state jurisdictions, not firms, for compliance with international oversight standards S-0061. States would back the standards by restricting imports from uncertified jurisdictions and by controlling exports of inputs such as specialised hardware S-0061.