Nintendo's HDL and source code also leaked from a hack a couple of years ago but given how litigious Nintendo is, nothing came out of it despite everyone expecting some emulators to pop op.
Obviously anyone copying the code verbatim or close to it deserves to get the lawyers coming after them, but plausible deniability should be strongly kept in mind for people doing RE+reimplement work anyway.
Maybe they can borrow some of the DLSS techniques
Hence why dlss 2.0 doesn't need per game training anymore. The "machine learning" part of it seems to be mostly killed off unless they somehow managed to train a fully generic model that can handle all art designs and game engine settings (eg, motion blur on vs. off). But it seems more likely it's a fancy generic take on temporal upscaling, possible influenced by ML but not dominated by it.
Which I guess is something we'll find out with this leak. How much AI is actually in dlss 2.0 and how much is "just" offloaded temporal super sampling (such as epic's TSR)