U.S. considers new AI software export control to China

The US government is considering AI software export restrictions to China in a series of efforts to prevent the use of advanced AI in military and cyberattacks.

AI chatbots resembling Molly Russell and Brianna Ghey were found on Character.ai, prompting public outcry.

The US government is considering new measures to limit China’s access to advanced artificial intelligence (AI) software. This initiative, driven by national security concerns, aims to prevent the use of these technologies in military applications and cyberattacks.

This potential measure follow broader US restrictions over export of AI chips and manufacturing tools to China. In the same context the US proposed a “know your customer” rule that  would require national cloud companies to inform the government when their services are used by foreign entities to train AI models that could potentially be deployed for cyberattacks. The new area of restriction aims to cover AI models and their core software.

The Biden administration’s proposal involves establishing regulatory controls over the export of proprietary or closed source AI models , which are developed and kept confidential by companies like OpenAI and Google DeepMind. Currently, nothing is stopping US AI giants, which have developed some of the most powerful closed source AI models, from selling them to almost anyone in the world without government oversight.

The Commerce Department is reportedly discussing the use of a computing power threshold, which was outlined in a recent AI executive order, to determine which AI models would be subject to export controls. This move is part of a broader effort to maintain technological superiority and manage the risks associated with AI advancements. The proposed controls would primarily target new models that have not yet been released, as existing technologies have not reached the defined thresholds.

These considerations come in response to the rapid development and potential misuse of AI technologies that could be used to enhance cyber and biological warfare capabilities. Recent discussions highlighted by researchers from Gryphon Scientific and the Rand Corporation emphasize that advanced AI models could assist in the development of biological weapons. Additionally, the Department of Homeland Security’s 2024 threat assessment warns that cyber actors are likely to leverage AI to conduct more sophisticated cyberattacks. The U.S. aims to establish a regulatory framework that can keep pace with technological advancements while addressing the complex challenges of effectively implementing export controls. The Commerce Department has yet to finalize any rules, indicating that the discussions are ongoing and that feedback from industry stakeholders will be essential in shaping the final regulatory approach.