Meta CEO Mark Zuckerberg sharply criticized the idea of blocking Chinese AI models in the United States. In his view, banning advanced developments from China is an ineffective tool that would only weaken American competitiveness. Instead of restrictive measures, he urged companies to systematically identify their own bottlenecks and invest in removing barriers in order to compete on equal footing with Chinese labs.

The Spark of Conflict: Kimi K3 from Moonshot AI

The statement came amid an escalating debate in Washington over possible restrictions on open-weight models. The trigger was the release of the Kimi K3 model from Beijing-based Moonshot AI — the world's first open model in the 3T class with 2.8 trillion parameters, native vision, and a context of 1 million tokens. According to the project's internal benchmarks, it is second only to proprietary giants — Claude Fable 5 from Anthropic and GPT 5.6 Sol from OpenAI.

Some U.S. officials and corporations suspect Chinese labs of using distillation — a method where one model is trained on the responses of another. In July, White House Office of Science and Technology Policy Director Michael Kratios directly accused Moonshot AI of this. Treasury Secretary Scott Bessent warned that if "hidden industrial distillation attacks" are identified, sanctions and inclusion in export restrictions will follow.

Industry Against Bans

Zuckerberg's position aligns with a large-scale campaign in defense of open weights. The Linux Foundation, together with Cisco, Dell Technologies, IBM, Meta, Microsoft, Nvidia, and 25 other companies, sent a letter urging not to restrict open-weight models. Nvidia CEO Jensen Huang also opposed bans, stating that high-quality Chinese open models should be used, and blocking them would harm the U.S. itself.

Simultaneously, the White House is forming a mechanism for voluntary pre-screening of the most powerful AI systems. According to a June decree, developers will be able to provide the government with access to frontier models 30 days before release. However, amid protests in San Francisco, where march participants demanded that OpenAI, Anthropic, and Google DeepMind suspend training of more powerful models, and former OpenAI researcher Daniel Kokotailo proposed freezing the race until 2040, it becomes clear: the industry is splitting into two camps.

My analysis: Zuckerberg's position is pragmatic — in the context of global competition, bans only stimulate the creation of closed ecosystems, which in the long term will weaken the U.S. innovation potential. Open models are not a threat but a catalyst for progress, and attempts to block them could lead to technological isolation.