The White House is developing a voluntary framework to vet advanced AI models, which is expected to exempt lower-cost, open-weight models. The policy aims to address national security risks while fostering American AI dominance, with a 30-day review period proposed for qualifying models.
This policy aims to balance national security concerns with fostering innovation in AI, potentially exempting widely adopted open-weight models while scrutinizing the most powerful systems. The approach could influence the global race for AI supremacy and impact the development and deployment of future AI technologies.
The White House is preparing a voluntary framework to vet advanced artificial intelligence models, with a plan to exempt lower-cost, open-weight models from the policy. This initiative follows a June 2 executive order aimed at mitigating the catastrophic risks associated with AI.
According to sources familiar with the plan, only "state-of-the-art" models identified as posing national security risks would be subject to the vetting process. This determination would be made through a benchmarking process managed by several federal agencies, including the National Security Agency.
The proposed policy also includes a maximum 30-day period for the government's review of new AI models. White House spokesperson Liz Huston stated that the voluntary framework supports the administration's cybersecurity strategy by strengthening national security and ensuring American AI dominance.
Industry support for open-weight models, whose core components are publicly available and customizable, has been widespread. Nvidia CEO Jensen Huang has been actively advocating on Capitol Hill against potential restrictions on these models. Skeptics have raised concerns that stringent safety regulations could inadvertently benefit Chinese companies and hinder U.S. leadership in advanced AI development.