Mark Zuckerberg has a message for Washington: banning Chinese AI models won’t work. In an interview with the Financial Times, the Meta CEO argued that restricting access to Chinese artificial intelligence technology is “not an effective solution” for maintaining American dominance in the AI race.
Instead of building walls, Zuckerberg wants to build better open-source models. And he’s not shy about pointing out who might benefit most from the restrictive approach: his direct competitors.
The regulatory capture argument
He specifically raised concerns about “regulatory capture” by US labs like OpenAI and Anthropic, the two companies most likely to benefit if the government restricts access to foreign AI alternatives.
Zuckerberg’s argument is that open models, ones that can be customized and run on local hardware, would actually address cybersecurity challenges faster than closed systems. It’s a counterintuitive claim that directly challenges the national security framing that proponents of Chinese AI bans typically rely on.
