To be a bit pedantic, your statement is a slightly ambiguous.
They have a good/accurate definition of open source. They don't "define it" in that they can change the definition to mean "and the packaging must be blue" and that becomes part of what open source means.
Whenever this debates comes up everyone always says, "wait, the term existed before OSI", but no, it really did not. Essentially nobody was saying "open source" before 1998 when OSI and Tim O'Reilly started pushing "open source".
https://thebaffler.com/salvos/the-meme-hustler
"Open source" wasn't our idea, and the greatest marketing trick that OSI ever did was convince so many of us that it was.
https://trademark.trademarkia.com/open-source-75439502.html
Bruce Perens regrets this[1]. But they could have. The term was not generic when they coined it. It really is like "kleenex".
----
[1] private communication
Simply put, I think if the name “open source” was a trademark, the open source community would still exist, and businesses would produce open source software as they do today too, but a lot of us would be using some other name for this instead of referring to it as “open source”, so that we’d avoid any potential problems with the trademark of open source.
https://groups.google.com/forum/#!msg/comp.os.ms-windows.pro...
http://www.xent.com/FoRK-archive/fall96/0269.html
Especially that Caldera one seems pretty high profile.
"Open source" has always been used as a generic term, because it's a very descriptive adjective+noun one. Trying to reserve such terms for one particular narrow usage seems like an fools' errand to me, which is why these discussion are being held all the time.
I didn't know about the NT example. It's a new one to me. But it's also very isolated. I said "essentially nobody" not absolutely "no one". There are a few scattered and rare examples before 1998, but it was not common parlance and most people had never heard about "open source" before Tim O'Reilly bankrolled OSI.