They could be used in bloody fast AI crawling and browse, for example dynamic webbased finetuning which involves non-static content: I think GET requests are safe in this matter.