Archive · 2 Sep 2026
Why America Should Not Imprison Frontier AI
America has started requiring licenses for some new AI releases.
The article says it probably will not stop regulating AI this way.
China is making powerful AI models that may be only months behind American ones.
Many Chinese models are open-weight, so people can use or change them.
One model called GLM 5.2 is said to be as good as older leading American models.
China may take longer to catch up with America’s newest model, Mythos, because it has fewer computer chips.
American companies are also trying to stop competitors from copying their models’ outputs for training.
The article says banning Chinese models might be hard to enforce and might not stop Chinese companies from improving them.
America has begun licensing AI releases, and the article says this policy is unlikely to stop.
China’s leading AI models are reportedly only months behind America’s best systems.
Many Chinese models are open-weight, allowing people to run or modify them.
GLM 5.2 from Z.ai reportedly matches the previous generation of leading American models.
The article argues that banning Chinese models would be difficult and might not stop China’s AI industry.
- Who
- The United States, China, American AI laboratories, and Chinese AI laboratories including Z.ai.
- What
- A debate over licensing, restricting, or banning frontier AI models, especially Chinese open-weight models.
- When
- Now; the article says Chinese models may be months behind American models, or at most about a year behind.
- Where
- The United States and China, with possible effects on foreign users of Chinese AI models.
- Why
- Because China’s leading models are rapidly approaching America’s, while open-weight systems can be widely used and modified, making restrictions difficult to enforce.
Restrict Frontier AI
Keep AI Open
Licensing AI releases
Restrict Frontier AI
Licensing can continue as a way for America to control the release of frontier AI systems.
Keep AI Open
The article argues that America should not effectively imprison frontier AI through excessive regulation.
Banning Chinese models
Restrict Frontier AI
America could ban Chinese open-weight models and punish foreign users in an effort to limit their spread.
Keep AI Open
Even an enforceable ban might not stop China’s model-makers because a large domestic market would continue supporting them.
Protecting American advantages
Restrict Frontier AI
Cracking down on model distillation could make it harder for competitors to train systems using outputs from the best American models.
Keep AI Open
The article says such measures may provide only months, or at most about a year, before Chinese models catch up.
- Current policy
- America has started licensing AI releases.
- China’s position
- China’s top models are described as only months behind America’s.
- Model openness
- Most of China’s top models are reportedly open-weight.
- Named model
- GLM 5.2, from Z.ai, is said to match the best American models from the previous generation.
- Newest American model
- Chinese laboratories may take longer to catch up with Mythos.
- Main constraint
- Chinese laboratories are described as having fewer chips.
- Proposed restriction
- America could ban Chinese open-weight models and penalize foreigners who use them.