That's good to know, thanks. In my case nearly 100% of my code ends up public on GitHub, so I assume everyone's code models are training on it anyway. But would be worth considering if I had proprietary codebases.