Only the American Chemical Society (ACS) does that.

He explicitly mentions an Elsevier database as the reason that he's worried about mining Elsevier data.

I think his post is effectively asking in a guess culture way. He wants that the publishers respond and say that their broad license doesn't limit what he's doing.

There is some of that in other posts on the blog. But that particular post is about images. He can't just click "agree" and do what he wants, but he has to explicitly ask them about every image project.