Analysts say fear of China is not the main blocker to AI rules

In short: A growing view in the US says the biggest obstacle to regulating advanced AI is domestic politics, not China racing ahead.

What's going on

In debates about regulating “frontier AI” (the most advanced AI systems), a common response is, “We cannot slow down because China will win.” On “The Ezra Klein Show,” AI policy researcher Matt Sheehan argues this China-focused story is often incomplete or wrong.

One reason is basic capacity. Recent reporting and analysis cited by The New York Times says China has far less computing power for training top AI models, roughly one-eighth to one-tenth of the US level. Computing power is like the size of an engine in a car, it limits how fast you can go.

Sheehan also points to China’s own AI rules. China has put in place binding regulations in recent years, including rules about recommendation systems, deepfakes, and labeling AI-made content. These rules focus heavily on controlling online information, but they show that a country can regulate and still have competitive AI companies.

What to watch

The harder problem may be agreement and enforcement. The US and China often do not even use the same definitions for terms like “AI safety” or “frontier model,” and there is little trust that any side would follow global rules. In the US and allied countries, the next key signal is whether leaders stop using “But China!” as a reason to avoid domestic steps like required testing or licensing (a permit to build or release the biggest models).

Source: NYTimes