Comment by zozbot234

The Chinese labs have been accused of training on elicited chat logs on a massive scale in violation of ToS. That's possibly a real concern, but it's nowhere close to "exfiltrating" the model or even roughly matching its behavior.