U.S. Steps Up Competition with China in Open-Source Models
The U.S. House Subcommittee on Commerce, Manufacturing, and Trade approved H.R. 10152, a bill aimed at taking global leadership in the open-source artificial intelligence field from China, and sent it to the full committee. The proposal, known as the Open-Source AI Leadership Act, seeks to expand the use of American-made open-weight models and inform the public about the risks associated with using Chinese competitors.
Chinese Models Doubled Their U.S. Rivals in Downloads
While leading American systems such as OpenAI’s GPT and Anthropic’s Claude remain closed-source, China-based technology companies have secured a strong position in the market by releasing open-weight models. Recent data show that Chinese-made open-source models were downloaded more than 328 million times worldwide in a single month. This figure is more than twice the total number of downloads of their American counterparts.
18 Months of Reporting and Coordination Responsibilities for the U.S. Department of Commerce
The approved H.R. 10152 bill provides for the establishment of a central point of contact within the U.S. Department of Commerce. The department will identify barriers to the use of American open-source models and develop partnerships with domestic companies, public institutions, and allied countries. During the 18-month period following the law’s enactment, the department will report data on censorship practices in Chinese-origin models, vulnerabilities in supply chains, and security risks.
This growing divergence in the technology sector is seen as an extension of the strategic initiatives pursued by the U.S. on a global scale. The Washington administration had launched a polarization process in AI diplomacy to influence other countries’ technology choices, sending letters to 35 countries and pressuring them to stay away from Beijing-based technologies.
An Approach That Prioritizes Security Without Imposing Market Restrictions
The bill specifically states that its provisions do not grant the authority to directly ban, restrict, or block open-source AI models available on the commercial market. While the bill aims to remove regulatory barriers and support U.S.-based developers, it also envisages monitoring models controlled by foreign adversary countries as a security risk.