Many of the open weight models are trained on outputs from these models (distillation)

If they provably shared those datasets then they're just as liable for piracy.

Most open models are developed for profit so they should be equally liable.