"Ecosystem Equals AI Leadership"...Opposition to Open AI Model Regulation Spreads

77 U.S. IT Companies, Including Microsoft, Nvidia, and Meta, Sign Open Letter
"Open AI Models Are the Key Foundation for Broadening AI Adoption"

U.S. information technology (IT) companies are joining the open-source artificial intelligence (AI) model advocacy movement recently initiated by Microsoft, Nvidia, and Meta. As the U.S. government seeks to contain China and tightens controls on the disclosure and overseas access of frontier (top-tier) AI models, these companies argue that the core of the AI supremacy race lies in expanding the ecosystem, rather than in achieving the most powerful models.

Microsoft, Nvidia, Meta, OpenAI, Google, and 77 other AI companies and investment firms sent an open letter to U.S. policymakers urging them to avoid excessive regulation of open-source models. Screenshot from the Microsoft website.

According to the industry on July 27, Microsoft, Nvidia, Meta, OpenAI, Google, and 77 other AI companies and investment firms have sent an open letter to U.S. policy authorities urging them to refrain from excessive regulation of open source models. The number of participating companies began at around 25 when the letter was first released, but quickly increased as major players including OpenAI and Google added their names.

Open Letter: "Open AI Models Drive Innovation"

The open letter emphasized that open-source models are essential for securing long-term U.S. AI sovereignty. The companies wrote, “Just as open-source software in the 1980s laid the foundation for the internet and the modern IT industry, AI leadership depends not on a single cutting-edge model but on a powerful and open ecosystem.” The letter further asserted, “Open AI models—ones that anyone can download, review, modify, and operate on their own infrastructure—are the critical bedrock for advancing innovation and competition, as well as for spreading AI more broadly.”

They also argued that open source can contribute to both AI safety and cybersecurity. Their reasoning is that closed models make it difficult for external researchers to identify errors or incidents of compromise, and if advanced technology is concentrated within a handful of companies, it becomes harder to guarantee technological transparency and safety. In contrast, releasing open-source model weights enables a diverse range of researchers and developers to analyze model behavior, identify vulnerabilities, and create new defensive measures.

The AI companies proposed that rather than regulating open-source models themselves, policymakers should develop measures focused on misuse cases. For example, they explained that “distillation” —using the output of one model to train another—should be classified as a legitimate model improvement technique, and should be distinguished from the illegal theft of intellectual property.

Jensen Huang, CEO of Nvidia, also personally shared the letter via his social networking service (SNS) account on X (formerly Twitter), stating, “AI is the technology that will transform every industry and will be built by both enterprises and nations,” and emphasized, “Both cutting-edge closed models and cutting-edge open models are needed.”

This movement comes at a time when China's pursuit in AI is accelerating and both the U.S. government and Congress are strengthening oversight and supervision of AI models. Triggered by a cyberattack involving the OpenAI GPT model, the U.S. Congress has proposed legislation requiring a mandatory “kill switch” for high-risk AI models. The U.S. government has also warned that sanctions could be imposed if evidence surfaces that AI technology has been misappropriated.

Meanwhile, Amazon Web Services (AWS) and Anthropic did not participate in this open letter. Anthropic has maintained its closed frontier model business structure, emphasizing AI safety concerns and noting that once model weights are released, it is difficult to recall them.

"Security Requires Closed Models, Ecosystem Growth Needs Open Models"...Korea Must Pursue a Two-Track Approach

Jensen Huang, CEO of Nvidia, is speaking at the San Francisco AI Summit on the 24th (local time). Photo by Yonhap News.

Experts have advised that Korea should also adopt a “two-track strategy”: deploying closed models in the national security sector to protect technology and data, while simultaneously nurturing open-source models for corporate applications and ecosystem development. Lee Jaeseong, a professor in the Department of AI at Chung-Ang University, explained, "To build an alternative ecosystem between the U.S. and China, a two-track strategy is inevitable: closed models for sovereign AI topics like defense and security, and open models for general applications.”

This content was produced with the assistance of AI translation services.