What's the best practice when using a copyrighted file for GPT?

I haven’t tested, but I don’t think it’s possible to download the PDF?

Note the last bit from the website in question who got the file names…