The article is about case reports, not about empirical studies. Putting a fake case report on GitHub wouldn't make it any less fake.

out of context that makes sense...but in the context of a case report how do you implement that? The patients have privacy rights and the authors/doctors have a responsibliity to protect them. That doesn't justify this but it does force a conversation about what 'every single data point' means. Does it mean the patient's real name and social security number? their complete medical chart?

Case reports are descriptive not determinative and should be treated as such by other scholars. They are 'I saw this' not 'this is generalizably true'. They can (and often are) replicated or countered but they are not per se research as you are thinking about it. Whether it is fictitious or not, other scholars should be cautious in citing them as proof/evidence in papers that fit into the 'research' mold.

And then there be large amounts of fake data for the next generation of AIs to learn from.

What is stopping anyone from faking the data they use in their research papers?

Sure it might be verifiable but if the data was made to give the desired results, i.e. faked to be what is required for the paper.