If it has already been established that open source doesn't mean open assets, why would we change that now? After all, training data is literally nothing but assets - except that you don't need them to run the application. So in that sense open LLMs are more open than these games.