The training of AI models never truly stops, with companies constantly training newer, bigger versions that are not yet released to the public.

AI companies that create closed ecosystems will generate massive amounts of training data from prompts, generated code, and production releases

For years, AI models could not be built very deep because they would be hard to train.

The original poster considered the possibility that one AI model was extensively trained using data from another.