@slothrop Funnily, this is actually neither. The plagiarized code is LGPL, the project it was submitted to is LGPL, and the robot even included the credit of the original author of the code. Somehow, it was so illegitimate that it underflowed back to being legitimate.

It is, however, an instructive example to those who claim that, pfft, an LLM’s training data is simply too large and diverse that it would ever just cough up GPL’d code directly