Trump team backs open AI models but debates limits on Chinese ones

In short: The Trump administration says it supports open AI models, but it is split on how to handle widely used Chinese open models over security worries.

What's going on

“Open source” or “open weight” AI models are systems that anyone can download, change, and share, including the core settings called “weights” (think of them like the model’s learned wiring). Because they are free to download, researchers, startups, and small businesses often use them instead of paying a single company.

In official policy, the administration has leaned toward supporting these open models. Its America’s AI Action Plan calls them valuable for innovation and says it should be up to developers to release a model openly or keep it closed. The plan also points to steps like helping small and medium-sized businesses adopt open models and improving access to expensive computing power, which is needed to train and run advanced AI.

The harder question is what to do about Chinese open models. Reporting says some U.S. agencies have discussed stronger warnings and possible restrictions, such as adding certain Chinese AI labs to an “Entity List” (a government list that can block U.S. access without a license). Other ideas included rules that would only allow U.S. companies to host Chinese models if they could prove the models are secure and accept responsibility if something goes wrong. Those proposals were reportedly blocked or softened due to concerns they could slow innovation.

What to watch

The current approach appears to avoid an outright ban while pushing security messaging about risks like hidden “backdoors” (secret ways in). Future China-focused actions are still possible, especially as more powerful Chinese models are expected to become freely downloadable.

Source: NYTimes