The reality is, passing this huge amount of data was the only way these crazy current AI models work as powerful as ChatGPT. With a restriction like you fantasize, the AI programs would have been dominated by bad actors and the west would not have a counter technology for a decade if not longer.
Regulating the outputs of AIs would be a separate story. But it's still overwhelmingly difficult. OpenAI is actually advanced in this region in the sense that they have in pocket the single best technology to politically balance the replies by a chatbot.