Microsoft, NVIDIA, IBM and other giants issued a joint statement: Strive to promote open weighted AI and consolidate the leading position of the United States.

Kuai Technology reported on July 25 that a total of 25 U.S. technology companies including Microsoft, Nvidia, Meta, and IBM jointly issued an open letter titled "Open Weight and U.S. AI Leadership," calling on the U.S. government to introduce supportive policies, relax controls, vigorously develop open weight AI models, and rely on open ecology to maintain the competitive advantage of the country's AI industry. The article compares open-weighted AI to the open source software movement of the 1980s and believes that open source software laid the foundation for the modern Internet industry. The current AI industry also needs an open model to drive industry expansion. The open weight model supports enterprises, start-ups, universities and public institutions to freely download, modify and independently deploy AI without having to build large models from scratch and bear high training costs. It can accelerate the implementation of AI in manufacturing, medical care, agriculture, education and other industries. At the same time, it fully activates market competition, reduces industry application costs, prevents the market from being monopolized by a single manufacturer, and allows enterprises to gain technological autonomy. The statement objectively acknowledges that there are hidden dangers in open weighting: after the model is made public, it is easily tampered with by third parties, and it is difficult for enterprises to fully control subsequent usage scenarios. However, companies oppose the direct ban on opening up models and advocate the establishment of a refined and differentiated regulatory framework. The core reason is that an open ecosystem can bring together a large number of R&D personnel around the world to jointly test and troubleshoot model security vulnerabilities, and continuously iterate protection mechanisms, which is more conducive to resisting network security risks than closed models. At the same time, it is clear to distinguish between legal model distillation optimization and illegal theft of technology for profit, and to oppose blanket restrictions on distillation technology. Participating companies unanimously believe that if the United States wants to maintain its leading position in the global AI field for a long time, it is not enough to rely solely on a few top closed-source models. It must build an open weighted industrial ecosystem covering the entire industry and find a balance between technological innovation, risk and safety, and industrial inclusiveness. The joint list shows that leading companies such as OpenAI, Google, and Anthropic, which specialize in large closed-source models, did not participate in signing the open letter.