Mark Zuckerberg opposes U.S. restrictions on Chinese AI models, advocating for open-weight models to enhance competitiveness and safety.
Mark Zuckerberg, the CEO of Meta, has expressed his belief that the United States should not impose restrictions on Chinese artificial intelligence models. He argues that such actions would not effectively address the ongoing tech rivalry and instead suggests that American firms should focus on improving their own capabilities.
This comment comes as discussions intensify in Washington about the possibility of limiting Chinese open-weight AI models, particularly following the launch of Kimi K3 by Moonshot AI, which claims to be the first open model of the 3T class with 2.8 trillion parameters, native vision, and a context span of 1 million tokens. This model is said to be competitive with proprietary systems like Claude Fable 5 and GPT 5.6 Sol.
Concerns Over Model Distillation
U.S. officials have raised alarms about potential intellectual property violations, with accusations that Chinese labs may engage in model distillation. In July, Michael Kratsios from the White House Office of Science and Technology Policy accused Moonshot AI of such practices, while Treasury Secretary Scott Bessent indicated that the government would closely monitor Chinese open-source models for any signs of theft.
Advocacy for Open Weights
Zuckerberg’s position aligns with a growing movement in favor of open-weight models. The Linux Foundation has backed a letter supporting this initiative, signed by major companies including Cisco, Dell Technologies, IBM, Microsoft, and Nvidia. The foundation emphasizes that openness in AI is crucial for ensuring safety, transparency, and reliability.
“Open-weight models provide the necessary visibility for researchers and enterprises to inspect, audit, and address vulnerabilities,” the Linux Foundation stated.
Nvidia’s CEO Jensen Huang also criticized the idea of banning beneficial Chinese models, warning that such actions could ultimately be detrimental to the U.S.
In response to these developments, the White House is working on a pre-screening mechanism for powerful AI systems, as outlined in a June directive aimed at allowing developers to provide access to their models prior to public release.
Additionally, in July, activists in San Francisco urged major AI companies to halt the training of more advanced models to prioritize safety, while former OpenAI researcher Daniel Kokotailo suggested pausing the AI race until 2040.
