They need to 1) release the model weights as open source, and 2) license code produced by the model as open source. It is a derivative work of open source code.

They need to 1) release the model weights as open source, and 2) license code produced by the model as open source. It is a derivative work of open source code. This is 100% a large corporation abusing and undermining the open source ecosystem. They can play by the rules or properly license a training dataset. People seem really eager to shill for a megacorp.