United States lawmakers from both parties have joined forces to push forward a bill that authorizes the U.S. government to implement export controls on artificial intelligence (AI) models to protect valuable American technology from foreign malicious actors.
The bill was introduced by Republican members of the House of Representatives, Michael McCaul and John Molenaar, along with Democratic Representatives Raja Krishnamoorthi and Susan Wild.
The legislation also specifically authorizes the Department of Commerce to prohibit Americans from collaborating with foreigners to develop AI systems that pose risks to U.S. national security.
This legislation aims to shield future AI export regulations from legal challenges and has also taken into account opinions from officials in the Biden administration during the drafting process.
This move also reflects Washington’s growing concern that adversaries like China and Russia could potentially utilize AI models to launch aggressive cyber-attacks or even create powerful biological weapons.
On Wednesday, sources informed Reuters that the United States is preparing to open a new front to restrict the use of generative AI developed by American companies by surrounding core software of advanced AI models like ChatGPT with fences.
However, under current U.S. laws, it would be significantly more difficult for the Department of Commerce, responsible for overseeing U.S. export policy, to regulate the export of freely available open-source AI models.
If this bill comes into effect, it would empower the Department of Commerce with explicit authority to regulate AI systems.
China has long heavily relied on many open-source models developed in the West, such as Meta Platforms’ “Llama.”
In March of this year, according to Chinese state media citing the Beijing Zhongyuan Artificial Intelligence Research Institute, most of China’s domestic AI models are actually built using Meta’s “Llama” model.
Once this bill is passed, it will pose a critical challenge to China’s AI development.