I've looked at the link provided by Amanda. As promised, I will not share the link, but I will comment on the contents.
Every result of any query I run has been published in the past. Each result is shown in XML format, and has a newspaper title, date of publication, city, state, and country of publication, etc, listed as variables in XML format. It appears as though there is a vulnerability on the hosting site that allows any person to run queries on their database by simply directing themselves to the appropriate URL.
This is simply a collection of older (and some newer) published articles, digitized using OCR technology.
Rereading your original post, I see that you only say the articles have not been published on the internet, not necessarily a claim of never being published in any medium. I did find that some of them have not been published on the internet before. I would caution anyone that reposting these on the internet would constitute a copyright violation, not only of the hosting site (who probably shares all of these to anyone with a membership), but also of the original publishers of the articles.
I indeed never claimed they were never seen before.. But they're not published on the internet, I haven't been able to find even one (which I tried to explain that guy that didnt listen.. being on the internet doesnt mean its published there)
Well, in my opinion, being on the internet = published. But, I see what you're getting at: the only place these articles are available is within a membership area, not in anything open to the general public. If the articles were to be "fixed up" by removing the OCR mistakes, they could be republished and indeed unique on the internet. However, it would still violate the copyright of the original publisher.
It's possible that one could get away with publishing the articles of obscure and out-of-business newspapers or magazines. If the publisher went out of business years ago, it's likely that no one from said publisher is looking for copyright violations.