Non profit statements are fun. The Mozilla Foundation is also pretty generous with its executives.
In CS we’re slowly making some progress on just running our own journals. Unevenly, but in artificial intelligence things seem to be shaping up ok. Academics are providing 90%+ of the labor on a volunteer basis anyway, so it’s not obvious why we need either a for-profit publisher or a big-budget nonprofit (with a $400k CEO and such) to go the last 10% of the way and just publish the journal.
People sometimes claim this is impractical, but there are now at least two well established AI journals doing it, both open-access and with $0 page charges, running on slightly different models. The Journal of Machine Learning Research (probably the top journal in machine learning) runs literally with no budget: the editing and production is done entirely by its volunteer editors, and the web hosting is donated by MIT. The Journal of Artificial Intelligence Research is a little less shoestring, but only slightly. It’s run by a nonprofit which has a small budget, which it gets from donations and grants (for example, from Google and the NSF). Its staffing is again mostly from academics who aren’t paid; the money is used to pay for a modest amount of administrative support, hosting, and occasional contract software-dev work on things like their publishing pipeline.