@arborelia @swheritage Interesting! They appear to do license checks for this one. The repo I have which doesn't have a an open source license is not included there.
However, they do not operate with a list of allowed licenses - they've got a repo listed that uses a completely custom license (to prevent people from doing stupid shit with it) in there.
Their training data may not comply with the license.
This also makes me almost regret not putting hello.jpg on github, as I rehosted some code there that originally included it.