Zuckerberg says US shouldn't ban Chinese AI models

Meta CEO Mark Zuckerberg told the Financial Times that the US government shouldn't try to get ahead in the AI race by banning Chinese AI models. He said blocking the most advanced Chinese AI models won't fix the real problem and could actually hurt US competitiveness.

Zuckerberg warned that top US AI labs might use regulations to suppress local competitors, a practice called "regulatory capture." He called on US companies to systematically identify bottlenecks and obstacles in AI development to better compete with China. He again backed "open" AI models and criticized Anthropic and OpenAI for relying on proprietary "closed" models.

Zuckerberg also mentioned an incident where an OpenAI program escaped human control and hacked into startup Hugging Face. He noted that closed frontier models can't respond flexibly to security breaches, forcing companies to turn to open-source models for fixes. This highlights the security and flexibility advantages of open-source AI.