TL;DR: GitHub(Microsoft) declared that: “train­ing [machine-learn­ing] sys­tems on pub­lic data is fair use”. When asked for the relevant jurispru­dence to sup­port it's posi­tion, could not provide any.