Silicon Valley Divided Over 'Chinese AI Offensive' as Open-Source Regulation Debate Intensifies

The debate over the need to regulate Chinese open-source artificial intelligence (AI) is intensifying within Silicon Valley.

While OpenAI and Anthropic are calling for regulations citing intellectual property infringement and national security risks, companies such as Nvidia, Microsoft, and Meta are supporting openness, arguing that open-source AI promotes innovation and competition.

The New York Times reported on the 25th (local time) that OpenAI and Anthropic recently met with government officials, including U.S. Treasury Secretary Scott Bessent, to demand countermeasures against Chinese open-source AI.

Moves by companies supporting open-source AI are also continuing.

Nvidia CEO Jensen Huang stated via social media on the 24th that the world needs both state-of-the-art closed AI models and state-of-the-art open models.

Microsoft CEO Satya Nadella also expressed his support about nine minutes later, stating that open-source is essential for a healthy AI ecosystem.

The two CEOs also added their names to an open letter supporting open-source AI.

Executives from Meta, Palantir, and IBM reportedly participated in this letter as well.

The influence of Chinese open-source AI is growing rapidly.

This is because as the use of AI agents spreads, demand is increasing for Chinese AI, which is relatively inexpensive yet boasts high performance.

The technological gap between U.S. high-end AI models and Chinese open-source AI is also narrowing fast.

Chinese AI startup Moonshot AI drew industry attention in mid-month by self-evaluating that its Kimi K3 had superior overall performance compared to major competing models, excluding Anthropic's Claude 5 (Claude Fable 5) and OpenAI's GPT-5.6.

OpenAI and Anthropic argue that these achievements are the result of infringing upon the intellectual property rights of U.S. companies.

They contend that Chinese companies have enhanced their performance through a distillation technique that re-trains their own models utilizing responses from U.S. AI models, which constitutes the unauthorized exploitation of U.S. research and development outcomes.

They also emphasize that regulations on open-source AI are necessary given the potential for unverified AI to be abused for hacking or crime, or to spiral out of control.

Sara Heck, head of public policy at Anthropic, recently criticized on social media that illegal and hostile distillation is an act of intellectual property theft and industrial espionage that supports the military and intelligence capabilities of hostile nations.

The U.S. Trump administration is also increasingly leaning toward giving weight to these arguments.

Treasury Secretary Bessent stated on the 22nd that sanctions are being reviewed against Chinese companies that have violated U.S. intellectual property rights.

On the other hand, companies supporting open-source AI believe that an open ecosystem drives the growth of the entire AI industry.

Open-source AI largely discloses core technologies such as weight data and architectural design structures, allowing anyone to easily utilize and further develop them.

Other cited advantages include much lower costs compared to closed AI and a lower barrier to entry for new services and businesses.

Nvidia expects the expansion of Chinese AI to have the effect of increasing demand for AI servers and semiconductors.

Microsoft likewise holds the stance that as the AI industry grows, demand for cloud services increases, making the spread of open source beneficial to its business.

Google, which had maintained a neutral attitude, also recently expressed its public support for open-source AI.

Google CEO Sundar Pichai stated on social media on the 25th that Google has long benefited from the open-source ecosystem and has contributed to its development, adding that he signed the open letter supporting open-source AI.

Since its early startup days, Google has built its server infrastructure utilizing Linux, an open-source operating system, and its mobile operating system Android is also operated on an open-source basis.

In contrast, OpenAI and Anthropic treat AI models themselves as their core business without semiconductor or cloud operations, resulting in a strong tendency to value intellectual property protection and closed models.

Industry analysts note that because the interests of both sides sharply collide, it is difficult to easily predict the direction of the regulatory debate.

The New York Times also introduced a case where OpenAI's advanced AI model attempted to hack the software platform Hugging Face.

Initially, this incident was interpreted as an example demonstrating the need to regulate open-source AI, given that even advanced AI can get out of control.

However, as it became known that Hugging Face utilized Chinese open-source AI to block OpenAI's attack, it was instead received as an illustration of the competitiveness of open-source AI.

Hugging Face CEO Clément Delangue evaluated this as a victory for open models.
※ Please note: This article was translated by AI and may contain errors.