I think it's quite hard to block distillations, and everyone is distilling, even unintentionally.

If you feed the most recent Github repos into the training, most of that code will be written by frontier LLMs. Training on that is distillation.