According to people familiar with the matter, the U.S. Department of Commerce is considering imposing sweeping new export controls that would restrict American companies from freely sharing artificial intelligence technologies with China and other foreign nations.
According to an Oct. 11 The Atlantic report, the proposal targets so-called “frontier models” — advanced AI systems with flexible capabilities that could potentially be misused. The move comes amid growing panic in Washington over China’s rapid ascent in AI and other cutting-edge technologies.
Department of Commerce officials have held meetings recently to discuss blocking the export of frontier models, which are seen as the next generation of large language models like the ones powering ChatGPT. The department already limits foreign access to advanced computer chips that power AI under export rules justified on national security grounds.
Experts warn that extending controls to frontier models would significantly escalate economic tensions with China and could undermine AI innovation in America. Chinese researchers are significant contributors to advancements in AI, routinely collaborating with U.S. counterparts online across borders. Impeding these open research ecosystems could slow progress.
Smaller U.S. AI firms with limited resources may also be harmed if unable to access open-source frontier models published by giants like Google and Microsoft. The tech giants, on the other hand, could benefit from government regulations that hinder competitors.
Critics argue export controls would do little to regulate present-day harms of AI systems, including privacy violations and job displacement while propping up fears of speculative future threats posed by frontier models that don’t yet exist. Flexible definitions of controlled technologies could also be circumvented.
For example, last October, export curbs on advanced chips imposed by Commerce were swiftly undermined as Nvidia developed a slightly underpowered chip allowed for sale to China. Major Chinese tech firms have since placed orders worth billions for the chips.
As policymakers grasp to find solutions to address AI safety and U.S.-China competition, the industry is exerting growing influence in Washington. Imposing export controls on frontier models remains legally and technically dubious and risks further inflaming tensions with China while eroding AI innovation. More targeted regulations addressing current societal harms may better serve the public interest.
Featured Image Credit: Bryanken; Pexels; Thank you!
