The US government is taking proactive measures to restrict the export of artificial intelligence (AI) models to Russia and China, citing concerns over potential risks such as deepfakes and biological weapons. A new law will empower the Biden administration to control AI exports, thereby slowing down China's progress in the field.

The Commerce Department is launching a compliance initiative to limit the export of open or closed-source AI models, shielding US interests and AI technology from potential misuse. This move complements previous efforts to block Chinese access to advanced computer chips.

US officials are particularly concerned about the core software of large language models, such as ChatGPT, falling into the wrong hands. Researchers warn that adversaries may exploit AI for cyber attacks and biological weapons development. The Chinese Embassy has opposed the move, labeling it "unilateral bullying and economic coercion."

Deepfakes pose a significant threat as a disinformation weapon, enabling realistic but fake videos that can manipulate public opinion. Social media platforms are already grappling with this issue, and the development of generative AI tools has made it easier for rogue actors to create and disseminate deepfakes.

Moreover, AI models can potentially leak information for developing biological weapons, a concern shared by researchers, intelligence agencies, and experts. The Department of Homeland Security also fears that AI can amplify cyberattacks on critical infrastructure.

To address these risks, a proposed bill aims to enforce controls on AI model exports, preventing potential adversaries from exploiting them. Experts emphasize the need for balanced regulations that foster innovation while mitigating AI-associated risks. Overly strict regulations could stifle progress and create a vacuum for overseas competitors, negatively impacting fields like infrastructure, national security, and drug discovery.