If he gets the data through Elsevier's system, he isn't really anonymous. And if he wants to do the particular project of this post, he has to get images, which he can't do just by clicking "agree," but he would have to tell them what he was going to do with the images. What's going to happen is that people are going to pirate all journal articles, largely for other reasons. At that point, he can do his project; I don't know whether anonymously. But this is a delay and a lot of wasted effort duplicating existing databases and query infrastructure.