But there's nothing to force me to provide you with the software until you've given me money. Even if I used someone else's GPL'ed software in my software, I can simply not distribute it until someone gives me $100,000.
"When you purchase access to an open-data paper"
Some provisos first. Some papers are gratis, and don't require being purchased. (Eg, author-pays papers.) Also, I think you are defining "open-data paper" that way, making your argument a bit tautological.
Those aside, I'm pointing out that the "open-data paper" movement started independently and likely co-temporally, if not earlier than, the free software movement.
Of course, you haven't infringed on anyone's freedom yet, there are no users. I guess the thing is if you massively overcharge without providing any corresponding service of value the first person that purchases it has the option to throw the source code up on the internet and it can be forked.
I guess you also have to find a buyer that is willing to pay you the large sum while also knowing they can't charge others for it without also handing over the source code, so it's doubtful anyone would do so.
There are other options though, if a single person or entity owns the copyright for the entire code base, with no outside contributors, they can dual license it. GPL for the public and then commercial proprietary license for businesses. That way the business can build on it if they wish without redistributing the source.
While, to go back to my first comment:
"If you have useful data, the free software principles say that you might as well charge a lot of money for it, which is not what most chemists and biologists think of in discussion of data openness."
Sure, there are distinctive features for each domain, but in broad strokes, I don't see the above description as very different from the one following:
When you purchase access to an open-data paper, the author is required to supply you also with the raw data, and you must be allowed to redistribute this (under terms and conditions.)