Yes, but by OpenAIs line of argument, the model itself isn’t piracy/theft/rights-infringing.

The output of the model might be, but that’s not the model creators problem. So by distributing the model, you’re no longer distributing infringing material.