Financial Times warns Chinese open-weight AI could shape global rules

In short: A Financial Times opinion piece argues that China’s push to share low-cost AI models could also export China’s standards and rules for how AI is run.

What's going on

China’s president Xi Jinping used a major AI conference in Shanghai to position China as a supporter of low-cost “open-source” AI, according to an opinion column by Lizzi Lee. In plain terms, this means sharing key parts of an AI model so more people can use it and build on it.

The column says many Chinese AI models are better described as “open weight,” not fully open source. “Weights” are the AI’s learned settings (like the knobs and dials inside a machine), and they are shared publicly. But the full “training data,” the information used to teach the AI, is not shared.

The author argues that open-weight models can be cheaper to adopt, and easier to customize for local needs. They can also help teams work around limits on access to the most advanced computer chips. The piece links this trend partly to US export controls, which raised costs for Chinese AI firms and pushed them to build more self-reliant software systems.

A key claim is geopolitical, not just technical. The column says that as more countries use Chinese AI models, they may gradually adopt Chinese technical standards and approaches to AI governance, meaning the rules for testing, safety, and oversight.

What to watch

The column notes a dilemma for safety. Once model weights are released, they can be copied and changed, and safety restrictions can be removed. Watch for “selective openness,” where some models stay widely available, but the most capable versions face extra limits or reviews.

Source: Financial Times