Foreign companies can be barred from selling infringing products in the United States.
Russian and Chinese consumers are less interested in English-language articles.
I can’t really get behind the argument that we need to let LLM companies use any material they want because other countries (with other languages, no less) might not have the same restrictions.
If you want some examples of LLMs held back by regulations, look into some of the examinations of how Chinese LLMs are clearly trained to avoid answering certain topics that their government deems sensitive.
But they're not; you can download open source Chinese base models like Yi and Deepseek and ask them about Tianmen Square yourself and see, they don't have any special filtering.
Isn't it just one additional step to automatically translate them?