As they don't mention it I doubt it.

Tabnine, a similar competitor, explicitly mentions this on their website:

" Tabnine only uses open-source code with permissive licenses for our Public Code trained AI model (MIT, Apache 2.0, BSD-2-Clause, BSD-3-Clause). "

Other commenters here say the completion quality is worse than Copilot. I use Tabnine for local short completions only and am quite happy with it. Didn't try Copilot yet.

You apparently can "opt out of public code" now. I didn't find an explanation for whether that properly limits it to permissive licenses though.

Update: It seems like they check whether the code it emits matches the training set and if it does it won't suggest it.

No they did not. You have to train on only fully permissive code to ensure that is not a problem