The US government should not block Chinese artificial intelligence in an attempt to gain the upper hand in the race to develop the groundbreaking technology, Meta chief executive Mark Zuckerberg has told the Financial Times. Zuckerberg warned that a US ban on Chinese AI models would not be “an effective solution” and suggested that US companies should instead “systematically” identify bottlenecks and roadblocks in order to better compete with Chinese AI firms.
His comments in an interview with the FT published Wednesday come as an intense debate rages in Washington over how the US should regulate the fast-developing technology and respond to increasingly sophisticated Chinese competition.
The Debate Over Chinese AI
Some senior Trump administration officials have claimed there is evidence that China’s Moonshot AI covertly used US rivals to train its latest model Kimi K3. Moonshot has denied the allegations. US Treasury Secretary Scott Bessent warned last week that “sanctions” could be considered if Chinese labs “cross the line into IP theft.”
The comments come just one day after the US government announced it would ban new humanoid and quadruped robot imports from foreign manufacturers, warning that they pose “unacceptable risks” to US national security. The ban also includes power converters used to connect renewable energy infrastructure and batteries to the power grid, the latest attempt by Washington to bolster and safeguard its domestic industries and supply chain amid an intensifying rivalry with China to dominate advanced technology and artificial intelligence.
The tension between Zuckerberg’s free-market stance and the more hawkish posture of the Trump administration reflects a genuine fault line within the US technology and policy establishment. On one side are those who believe that open competition, including the free flow of open-source AI models, ultimately benefits American innovation. On the other are those who see Chinese AI as a national security threat that warrants aggressive countermeasures.
Meta’s Open-Source Strategy and Its Complications
Zuckerberg’s comments align with Meta’s broader strategy of championing open-source AI development. By releasing its Llama models with open weights, Meta has positioned itself as a counterweight to the closed, proprietary approaches of rivals like OpenAI and Anthropic. This open-source philosophy, however, complicates the regulatory landscape. If the US were to ban Chinese AI models, it would raise difficult questions about how to enforce such a ban in an ecosystem where open-source weights can be freely downloaded and modified globally.
The broader context is important. As EastFrontier has previously reported, China’s open-source AI models now rank first globally in downloads, and the co-opetition reality between China and the US in AI development is deeply entangled. A blanket ban would be extraordinarily difficult to enforce and could push Chinese AI development further underground, making it harder to monitor and assess.
Competition Over Restriction
Instead of relying on bans, Zuckerberg’s call for US companies to “systematically” address bottlenecks suggests a focus on improving domestic infrastructure, talent pipelines, and compute efficiency to maintain a competitive edge. This approach mirrors the philosophy behind the CHIPS and Science Act, investing in American strengths rather than simply trying to constrain Chinese ones.
Furthermore, Zuckerberg’s warning highlights a growing concern among some tech leaders that overly aggressive regulatory actions aimed at China could inadvertently stifle innovation and create a fragmented global AI ecosystem. The debate over how to manage the rise of Chinese AI models will likely intensify in the months ahead, particularly as the September Xi-Trump summit approaches and both sides weigh the costs and benefits of escalation versus accommodation in the AI domain.
The stakes of this debate extend well beyond the US-China bilateral relationship. If Washington were to ban Chinese AI models, it would need to decide whether to extend that ban to allied nations, a move that would almost certainly face resistance from European and Asian partners who have already integrated Chinese AI tools into research and commercial workflows. The alternative, a US-only ban, would create an asymmetric environment in which American companies compete under tighter constraints than their international peers. Zuckerberg’s intervention is therefore not just a defense of Meta’s open-source strategy; it is an argument about the structural conditions under which American AI can remain globally dominant.
The timing of his remarks is also significant. Meta is in the middle of its own aggressive AI investment cycle, with capital expenditure guidance raised to $64-72 billion for 2026. The company has a direct commercial interest in a world where AI models, including Chinese ones, circulate freely, because open competition validates Meta’s open-source bet and keeps the market from consolidating around closed incumbents. Zuckerberg is simultaneously a disinterested voice for free markets and a deeply interested party in the outcome of the regulatory debate. Washington will need to weigh both dimensions as it decides how to respond.
What is clear is that the window for a simple legislative fix is narrowing. Chinese AI models are already embedded in research workflows, developer toolchains, and enterprise pilots across the US and its allies. Any ban would require not just a prohibition on new downloads but a mechanism for enforcing removal of existing deployments, a technically and politically complex undertaking that would likely face immediate legal challenges. Zuckerberg’s intervention may not change the outcome of the policy debate, but it has raised the cost of the most aggressive option.