This is literally an admission that the dataset is strictly open source code with permissive licenses. According to the paper jetbrains wrote about full line completion they actually further processed the datasets used to ensure only permissively licensed code was used.
