[omegacharlie]

Think some of the negativity about Copilot may be the perception that if an individual or small startup attempted training an ML model from public source-code and commercialised a service from it they would be drowning in legal issues from big companies not happy with their code used in such a product.

In addition just because code is available publicly on GitHub does not necessarily mean it is permissively licensed to use elsewhere, even with attribution. Copyright holders not happy with their copyrighted works publicly accessible can use the DMCA to issue take-downs that GitHub does comply with but how that interacts with Copilot and any of its training data is a different question.

As much as the DMCA is bad law rather funny seeing Microsoft be charged in this lawsuit with the less known provision against 'removal of copyright management information'. Microsoft does have more resources to mount at defence so it will probably end up different compared to a smaller player facing this action.